drivers/net/ethernet/ibm/ibmvnic.c

   1 /**************************************************************************/
   2 /*                                                                        */
   3 /*  IBM System i and System p Virtual NIC Device Driver                   */
   4 /*  Copyright (C) 2014 IBM Corp.                                          */
   5 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
   6 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
   7 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
   8 /*                                                                        */
   9 /*  This program is free software; you can redistribute it and/or modify  */
  10 /*  it under the terms of the GNU General Public License as published by  */
  11 /*  the Free Software Foundation; either version 2 of the License, or     */
  12 /*  (at your option) any later version.                                   */
  13 /*                                                                        */
  14 /*  This program is distributed in the hope that it will be useful,       */
  15 /*  but WITHOUT ANY WARRANTY; without even the implied warranty of        */
  16 /*  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         */
  17 /*  GNU General Public License for more details.                          */
  18 /*                                                                        */
  19 /*  You should have received a copy of the GNU General Public License     */
  20 /*  along with this program.                                              */
  21 /*                                                                        */
  22 /* This module contains the implementation of a virtual ethernet device   */
  23 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
  24 /* option of the RS/6000 Platform Architecture to interface with virtual  */
  25 /* ethernet NICs that are presented to the partition by the hypervisor.   */
  26 /*                                                                         */
  27 /* Messages are passed between the VNIC driver and the VNIC server using  */
  28 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
  29 /* issue and receive commands that initiate communication with the server */
  30 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
  31 /* are used by the driver to notify the server that a packet is           */
  32 /* ready for transmission or that a buffer has been added to receive a    */
  33 /* packet. Subsequently, sCRQs are used by the server to notify the       */
  34 /* driver that a packet transmission has been completed or that a packet  */
  35 /* has been received and placed in a waiting buffer.                      */
  36 /*                                                                        */
  37 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
  38 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
  39 /* or receive has been completed, the VNIC driver is required to use      */
  40 /* "long term mapping". This entails that large, continuous DMA mapped    */
  41 /* buffers are allocated on driver initialization and these buffers are   */
  42 /* then continuously reused to pass skbs to and from the VNIC server.     */
  43 /*                                                                        */
  44 /**************************************************************************/
  45
  46 #include <linux/module.h>
  47 #include <linux/moduleparam.h>
  48 #include <linux/types.h>
  49 #include <linux/errno.h>
  50 #include <linux/completion.h>
  51 #include <linux/ioport.h>
  52 #include <linux/dma-mapping.h>
  53 #include <linux/kernel.h>
  54 #include <linux/netdevice.h>
  55 #include <linux/etherdevice.h>
  56 #include <linux/skbuff.h>
  57 #include <linux/init.h>
  58 #include <linux/delay.h>
  59 #include <linux/mm.h>
  60 #include <linux/ethtool.h>
  61 #include <linux/proc_fs.h>
  62 #include <linux/if_arp.h>
  63 #include <linux/in.h>
  64 #include <linux/ip.h>
  65 #include <linux/ipv6.h>
  66 #include <linux/irq.h>
  67 #include <linux/kthread.h>
  68 #include <linux/seq_file.h>
  69 #include <linux/interrupt.h>
  70 #include <net/net_namespace.h>
  71 #include <asm/hvcall.h>
  72 #include <linux/atomic.h>
  73 #include <asm/vio.h>
  74 #include <asm/iommu.h>
  75 #include <linux/uaccess.h>
  76 #include <asm/firmware.h>
  77 #include <linux/workqueue.h>
  78 #include <linux/if_vlan.h>
  79 #include <linux/utsname.h>
  80
  81 #include "ibmvnic.h"
  82
  83 static const char ibmvnic_driver_name[] = "ibmvnic";
  84 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
  85
  86 MODULE_AUTHOR("Santiago Leon");
  87 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
  88 MODULE_LICENSE("GPL");
  89 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
  90
  91 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
  92 static int ibmvnic_remove(struct vio_dev *);
  93 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
  94 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
  95 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
  96 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
  97 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
  98 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
  99                        union sub_crq *sub_crq);
 100 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
 101 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
 102 static int enable_scrq_irq(struct ibmvnic_adapter *,
 103                            struct ibmvnic_sub_crq_queue *);
 104 static int disable_scrq_irq(struct ibmvnic_adapter *,
 105                             struct ibmvnic_sub_crq_queue *);
 106 static int pending_scrq(struct ibmvnic_adapter *,
 107                         struct ibmvnic_sub_crq_queue *);
 108 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
 109                                         struct ibmvnic_sub_crq_queue *);
 110 static int ibmvnic_poll(struct napi_struct *napi, int data);
 111 static void send_map_query(struct ibmvnic_adapter *adapter);
 112 static int send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
 113 static int send_request_unmap(struct ibmvnic_adapter *, u8);
 114 static int send_login(struct ibmvnic_adapter *adapter);
 115 static void send_cap_queries(struct ibmvnic_adapter *adapter);
 116 static int init_sub_crqs(struct ibmvnic_adapter *);
 117 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
 118 static int ibmvnic_init(struct ibmvnic_adapter *);
 119 static int ibmvnic_reset_init(struct ibmvnic_adapter *);
 120 static void release_crq_queue(struct ibmvnic_adapter *);
 121 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p);
 122 static int init_crq_queue(struct ibmvnic_adapter *adapter);
 123
 124 struct ibmvnic_stat {
 125         char name[ETH_GSTRING_LEN];
 126         int offset;
 127 };
 128
 129 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
 130                              offsetof(struct ibmvnic_statistics, stat))
 131 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
 132
 133 static const struct ibmvnic_stat ibmvnic_stats[] = {
 134         {"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
 135         {"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
 136         {"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
 137         {"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
 138         {"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
 139         {"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
 140         {"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
 141         {"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
 142         {"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
 143         {"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
 144         {"align_errors", IBMVNIC_STAT_OFF(align_errors)},
 145         {"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
 146         {"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
 147         {"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
 148         {"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
 149         {"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
 150         {"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
 151         {"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
 152         {"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
 153         {"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
 154         {"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
 155         {"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
 156 };
 157
 158 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
 159                           unsigned long length, unsigned long *number,
 160                           unsigned long *irq)
 161 {
 162         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
 163         long rc;
 164
 165         rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
 166         *number = retbuf[0];
 167         *irq = retbuf[1];
 168
 169         return rc;
 170 }
 171
 172 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
 173                                 struct ibmvnic_long_term_buff *ltb, int size)
 174 {
 175         struct device *dev = &adapter->vdev->dev;
 176         int rc;
 177
 178         ltb->size = size;
 179         ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
 180                                        GFP_KERNEL);
 181
 182         if (!ltb->buff) {
 183                 dev_err(dev, "Couldn't alloc long term buffer\n");
 184                 return -ENOMEM;
 185         }
 186         ltb->map_id = adapter->map_id;
 187         adapter->map_id++;
 188
 189         init_completion(&adapter->fw_done);
 190         rc = send_request_map(adapter, ltb->addr,
 191                               ltb->size, ltb->map_id);
 192         if (rc) {
 193                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 194                 return rc;
 195         }
 196         wait_for_completion(&adapter->fw_done);
 197
 198         if (adapter->fw_done_rc) {
 199                 dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
 200                         adapter->fw_done_rc);
 201                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 202                 return -1;
 203         }
 204         return 0;
 205 }
 206
 207 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
 208                                 struct ibmvnic_long_term_buff *ltb)
 209 {
 210         struct device *dev = &adapter->vdev->dev;
 211
 212         if (!ltb->buff)
 213                 return;
 214
 215         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
 216             adapter->reset_reason != VNIC_RESET_MOBILITY)
 217                 send_request_unmap(adapter, ltb->map_id);
 218         dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 219 }
 220
 221 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
 222                                 struct ibmvnic_long_term_buff *ltb)
 223 {
 224         int rc;
 225
 226         memset(ltb->buff, 0, ltb->size);
 227
 228         init_completion(&adapter->fw_done);
 229         rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
 230         if (rc)
 231                 return rc;
 232         wait_for_completion(&adapter->fw_done);
 233
 234         if (adapter->fw_done_rc) {
 235                 dev_info(&adapter->vdev->dev,
 236                          "Reset failed, attempting to free and reallocate buffer\n");
 237                 free_long_term_buff(adapter, ltb);
 238                 return alloc_long_term_buff(adapter, ltb, ltb->size);
 239         }
 240         return 0;
 241 }
 242
 243 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
 244 {
 245         int i;
 246
 247         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 248              i++)
 249                 adapter->rx_pool[i].active = 0;
 250 }
 251
 252 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
 253                               struct ibmvnic_rx_pool *pool)
 254 {
 255         int count = pool->size - atomic_read(&pool->available);
 256         struct device *dev = &adapter->vdev->dev;
 257         int buffers_added = 0;
 258         unsigned long lpar_rc;
 259         union sub_crq sub_crq;
 260         struct sk_buff *skb;
 261         unsigned int offset;
 262         dma_addr_t dma_addr;
 263         unsigned char *dst;
 264         u64 *handle_array;
 265         int shift = 0;
 266         int index;
 267         int i;
 268
 269         if (!pool->active)
 270                 return;
 271
 272         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 273                                       be32_to_cpu(adapter->login_rsp_buf->
 274                                       off_rxadd_subcrqs));
 275
 276         for (i = 0; i < count; ++i) {
 277                 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
 278                 if (!skb) {
 279                         dev_err(dev, "Couldn't replenish rx buff\n");
 280                         adapter->replenish_no_mem++;
 281                         break;
 282                 }
 283
 284                 index = pool->free_map[pool->next_free];
 285
 286                 if (pool->rx_buff[index].skb)
 287                         dev_err(dev, "Inconsistent free_map!\n");
 288
 289                 /* Copy the skb to the long term mapped DMA buffer */
 290                 offset = index * pool->buff_size;
 291                 dst = pool->long_term_buff.buff + offset;
 292                 memset(dst, 0, pool->buff_size);
 293                 dma_addr = pool->long_term_buff.addr + offset;
 294                 pool->rx_buff[index].data = dst;
 295
 296                 pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
 297                 pool->rx_buff[index].dma = dma_addr;
 298                 pool->rx_buff[index].skb = skb;
 299                 pool->rx_buff[index].pool_index = pool->index;
 300                 pool->rx_buff[index].size = pool->buff_size;
 301
 302                 memset(&sub_crq, 0, sizeof(sub_crq));
 303                 sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
 304                 sub_crq.rx_add.correlator =
 305                     cpu_to_be64((u64)&pool->rx_buff[index]);
 306                 sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
 307                 sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
 308
 309                 /* The length field of the sCRQ is defined to be 24 bits so the
 310                  * buffer size needs to be left shifted by a byte before it is
 311                  * converted to big endian to prevent the last byte from being
 312                  * truncated.
 313                  */
 314 #ifdef __LITTLE_ENDIAN__
 315                 shift = 8;
 316 #endif
 317                 sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
 318
 319                 lpar_rc = send_subcrq(adapter, handle_array[pool->index],
 320                                       &sub_crq);
 321                 if (lpar_rc != H_SUCCESS)
 322                         goto failure;
 323
 324                 buffers_added++;
 325                 adapter->replenish_add_buff_success++;
 326                 pool->next_free = (pool->next_free + 1) % pool->size;
 327         }
 328         atomic_add(buffers_added, &pool->available);
 329         return;
 330
 331 failure:
 332         if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
 333                 dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
 334         pool->free_map[pool->next_free] = index;
 335         pool->rx_buff[index].skb = NULL;
 336
 337         dev_kfree_skb_any(skb);
 338         adapter->replenish_add_buff_failure++;
 339         atomic_add(buffers_added, &pool->available);
 340
 341         if (lpar_rc == H_CLOSED || adapter->failover_pending) {
 342                 /* Disable buffer pool replenishment and report carrier off if
 343                  * queue is closed or pending failover.
 344                  * Firmware guarantees that a signal will be sent to the
 345                  * driver, triggering a reset.
 346                  */
 347                 deactivate_rx_pools(adapter);
 348                 netif_carrier_off(adapter->netdev);
 349         }
 350 }
 351
 352 static void replenish_pools(struct ibmvnic_adapter *adapter)
 353 {
 354         int i;
 355
 356         adapter->replenish_task_cycles++;
 357         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 358              i++) {
 359                 if (adapter->rx_pool[i].active)
 360                         replenish_rx_pool(adapter, &adapter->rx_pool[i]);
 361         }
 362 }
 363
 364 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
 365 {
 366         kfree(adapter->tx_stats_buffers);
 367         kfree(adapter->rx_stats_buffers);
 368         adapter->tx_stats_buffers = NULL;
 369         adapter->rx_stats_buffers = NULL;
 370 }
 371
 372 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
 373 {
 374         adapter->tx_stats_buffers =
 375                                 kcalloc(IBMVNIC_MAX_QUEUES,
 376                                         sizeof(struct ibmvnic_tx_queue_stats),
 377                                         GFP_KERNEL);
 378         if (!adapter->tx_stats_buffers)
 379                 return -ENOMEM;
 380
 381         adapter->rx_stats_buffers =
 382                                 kcalloc(IBMVNIC_MAX_QUEUES,
 383                                         sizeof(struct ibmvnic_rx_queue_stats),
 384                                         GFP_KERNEL);
 385         if (!adapter->rx_stats_buffers)
 386                 return -ENOMEM;
 387
 388         return 0;
 389 }
 390
 391 static void release_stats_token(struct ibmvnic_adapter *adapter)
 392 {
 393         struct device *dev = &adapter->vdev->dev;
 394
 395         if (!adapter->stats_token)
 396                 return;
 397
 398         dma_unmap_single(dev, adapter->stats_token,
 399                          sizeof(struct ibmvnic_statistics),
 400                          DMA_FROM_DEVICE);
 401         adapter->stats_token = 0;
 402 }
 403
 404 static int init_stats_token(struct ibmvnic_adapter *adapter)
 405 {
 406         struct device *dev = &adapter->vdev->dev;
 407         dma_addr_t stok;
 408
 409         stok = dma_map_single(dev, &adapter->stats,
 410                               sizeof(struct ibmvnic_statistics),
 411                               DMA_FROM_DEVICE);
 412         if (dma_mapping_error(dev, stok)) {
 413                 dev_err(dev, "Couldn't map stats buffer\n");
 414                 return -1;
 415         }
 416
 417         adapter->stats_token = stok;
 418         netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
 419         return 0;
 420 }
 421
 422 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
 423 {
 424         struct ibmvnic_rx_pool *rx_pool;
 425         int rx_scrqs;
 426         int i, j, rc;
 427         u64 *size_array;
 428
 429         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 430                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 431
 432         rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 433         for (i = 0; i < rx_scrqs; i++) {
 434                 rx_pool = &adapter->rx_pool[i];
 435
 436                 netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
 437
 438                 if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
 439                         free_long_term_buff(adapter, &rx_pool->long_term_buff);
 440                         rx_pool->buff_size = be64_to_cpu(size_array[i]);
 441                         alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 442                                              rx_pool->size *
 443                                              rx_pool->buff_size);
 444                 } else {
 445                         rc = reset_long_term_buff(adapter,
 446                                                   &rx_pool->long_term_buff);
 447                 }
 448
 449                 if (rc)
 450                         return rc;
 451
 452                 for (j = 0; j < rx_pool->size; j++)
 453                         rx_pool->free_map[j] = j;
 454
 455                 memset(rx_pool->rx_buff, 0,
 456                        rx_pool->size * sizeof(struct ibmvnic_rx_buff));
 457
 458                 atomic_set(&rx_pool->available, 0);
 459                 rx_pool->next_alloc = 0;
 460                 rx_pool->next_free = 0;
 461                 rx_pool->active = 1;
 462         }
 463
 464         return 0;
 465 }
 466
 467 static void release_rx_pools(struct ibmvnic_adapter *adapter)
 468 {
 469         struct ibmvnic_rx_pool *rx_pool;
 470         int i, j;
 471
 472         if (!adapter->rx_pool)
 473                 return;
 474
 475         for (i = 0; i < adapter->num_active_rx_pools; i++) {
 476                 rx_pool = &adapter->rx_pool[i];
 477
 478                 netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
 479
 480                 kfree(rx_pool->free_map);
 481                 free_long_term_buff(adapter, &rx_pool->long_term_buff);
 482
 483                 if (!rx_pool->rx_buff)
 484                         continue;
 485
 486                 for (j = 0; j < rx_pool->size; j++) {
 487                         if (rx_pool->rx_buff[j].skb) {
 488                                 dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
 489                                 rx_pool->rx_buff[j].skb = NULL;
 490                         }
 491                 }
 492
 493                 kfree(rx_pool->rx_buff);
 494         }
 495
 496         kfree(adapter->rx_pool);
 497         adapter->rx_pool = NULL;
 498         adapter->num_active_rx_pools = 0;
 499 }
 500
 501 static int init_rx_pools(struct net_device *netdev)
 502 {
 503         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 504         struct device *dev = &adapter->vdev->dev;
 505         struct ibmvnic_rx_pool *rx_pool;
 506         int rxadd_subcrqs;
 507         u64 *size_array;
 508         int i, j;
 509
 510         rxadd_subcrqs =
 511                 be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 512         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 513                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 514
 515         adapter->rx_pool = kcalloc(rxadd_subcrqs,
 516                                    sizeof(struct ibmvnic_rx_pool),
 517                                    GFP_KERNEL);
 518         if (!adapter->rx_pool) {
 519                 dev_err(dev, "Failed to allocate rx pools\n");
 520                 return -1;
 521         }
 522
 523         adapter->num_active_rx_pools = rxadd_subcrqs;
 524
 525         for (i = 0; i < rxadd_subcrqs; i++) {
 526                 rx_pool = &adapter->rx_pool[i];
 527
 528                 netdev_dbg(adapter->netdev,
 529                            "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
 530                            i, adapter->req_rx_add_entries_per_subcrq,
 531                            be64_to_cpu(size_array[i]));
 532
 533                 rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
 534                 rx_pool->index = i;
 535                 rx_pool->buff_size = be64_to_cpu(size_array[i]);
 536                 rx_pool->active = 1;
 537
 538                 rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
 539                                             GFP_KERNEL);
 540                 if (!rx_pool->free_map) {
 541                         release_rx_pools(adapter);
 542                         return -1;
 543                 }
 544
 545                 rx_pool->rx_buff = kcalloc(rx_pool->size,
 546                                            sizeof(struct ibmvnic_rx_buff),
 547                                            GFP_KERNEL);
 548                 if (!rx_pool->rx_buff) {
 549                         dev_err(dev, "Couldn't alloc rx buffers\n");
 550                         release_rx_pools(adapter);
 551                         return -1;
 552                 }
 553
 554                 if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 555                                          rx_pool->size * rx_pool->buff_size)) {
 556                         release_rx_pools(adapter);
 557                         return -1;
 558                 }
 559
 560                 for (j = 0; j < rx_pool->size; ++j)
 561                         rx_pool->free_map[j] = j;
 562
 563                 atomic_set(&rx_pool->available, 0);
 564                 rx_pool->next_alloc = 0;
 565                 rx_pool->next_free = 0;
 566         }
 567
 568         return 0;
 569 }
 570
 571 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
 572                              struct ibmvnic_tx_pool *tx_pool)
 573 {
 574         int rc, i;
 575
 576         rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
 577         if (rc)
 578                 return rc;
 579
 580         memset(tx_pool->tx_buff, 0,
 581                tx_pool->num_buffers *
 582                sizeof(struct ibmvnic_tx_buff));
 583
 584         for (i = 0; i < tx_pool->num_buffers; i++)
 585                 tx_pool->free_map[i] = i;
 586
 587         tx_pool->consumer_index = 0;
 588         tx_pool->producer_index = 0;
 589
 590         return 0;
 591 }
 592
 593 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
 594 {
 595         int tx_scrqs;
 596         int i, rc;
 597
 598         tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 599         for (i = 0; i < tx_scrqs; i++) {
 600                 rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
 601                 if (rc)
 602                         return rc;
 603                 rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
 604                 if (rc)
 605                         return rc;
 606         }
 607
 608         return 0;
 609 }
 610
 611 static void release_vpd_data(struct ibmvnic_adapter *adapter)
 612 {
 613         if (!adapter->vpd)
 614                 return;
 615
 616         kfree(adapter->vpd->buff);
 617         kfree(adapter->vpd);
 618
 619         adapter->vpd = NULL;
 620 }
 621
 622 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
 623                                 struct ibmvnic_tx_pool *tx_pool)
 624 {
 625         kfree(tx_pool->tx_buff);
 626         kfree(tx_pool->free_map);
 627         free_long_term_buff(adapter, &tx_pool->long_term_buff);
 628 }
 629
 630 static void release_tx_pools(struct ibmvnic_adapter *adapter)
 631 {
 632         int i;
 633
 634         if (!adapter->tx_pool)
 635                 return;
 636
 637         for (i = 0; i < adapter->num_active_tx_pools; i++) {
 638                 release_one_tx_pool(adapter, &adapter->tx_pool[i]);
 639                 release_one_tx_pool(adapter, &adapter->tso_pool[i]);
 640         }
 641
 642         kfree(adapter->tx_pool);
 643         adapter->tx_pool = NULL;
 644         kfree(adapter->tso_pool);
 645         adapter->tso_pool = NULL;
 646         adapter->num_active_tx_pools = 0;
 647 }
 648
 649 static int init_one_tx_pool(struct net_device *netdev,
 650                             struct ibmvnic_tx_pool *tx_pool,
 651                             int num_entries, int buf_size)
 652 {
 653         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 654         int i;
 655
 656         tx_pool->tx_buff = kcalloc(num_entries,
 657                                    sizeof(struct ibmvnic_tx_buff),
 658                                    GFP_KERNEL);
 659         if (!tx_pool->tx_buff)
 660                 return -1;
 661
 662         if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
 663                                  num_entries * buf_size))
 664                 return -1;
 665
 666         tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
 667         if (!tx_pool->free_map)
 668                 return -1;
 669
 670         for (i = 0; i < num_entries; i++)
 671                 tx_pool->free_map[i] = i;
 672
 673         tx_pool->consumer_index = 0;
 674         tx_pool->producer_index = 0;
 675         tx_pool->num_buffers = num_entries;
 676         tx_pool->buf_size = buf_size;
 677
 678         return 0;
 679 }
 680
 681 static int init_tx_pools(struct net_device *netdev)
 682 {
 683         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 684         int tx_subcrqs;
 685         int i, rc;
 686
 687         tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 688         adapter->tx_pool = kcalloc(tx_subcrqs,
 689                                    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 690         if (!adapter->tx_pool)
 691                 return -1;
 692
 693         adapter->tso_pool = kcalloc(tx_subcrqs,
 694                                     sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 695         if (!adapter->tso_pool)
 696                 return -1;
 697
 698         adapter->num_active_tx_pools = tx_subcrqs;
 699
 700         for (i = 0; i < tx_subcrqs; i++) {
 701                 rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
 702                                       adapter->req_tx_entries_per_subcrq,
 703                                       adapter->req_mtu + VLAN_HLEN);
 704                 if (rc) {
 705                         release_tx_pools(adapter);
 706                         return rc;
 707                 }
 708
 709                 init_one_tx_pool(netdev, &adapter->tso_pool[i],
 710                                  IBMVNIC_TSO_BUFS,
 711                                  IBMVNIC_TSO_BUF_SZ);
 712                 if (rc) {
 713                         release_tx_pools(adapter);
 714                         return rc;
 715                 }
 716         }
 717
 718         return 0;
 719 }
 720
 721 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
 722 {
 723         int i;
 724
 725         if (adapter->napi_enabled)
 726                 return;
 727
 728         for (i = 0; i < adapter->req_rx_queues; i++)
 729                 napi_enable(&adapter->napi[i]);
 730
 731         adapter->napi_enabled = true;
 732 }
 733
 734 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
 735 {
 736         int i;
 737
 738         if (!adapter->napi_enabled)
 739                 return;
 740
 741         for (i = 0; i < adapter->req_rx_queues; i++) {
 742                 netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
 743                 napi_disable(&adapter->napi[i]);
 744         }
 745
 746         adapter->napi_enabled = false;
 747 }
 748
 749 static int init_napi(struct ibmvnic_adapter *adapter)
 750 {
 751         int i;
 752
 753         adapter->napi = kcalloc(adapter->req_rx_queues,
 754                                 sizeof(struct napi_struct), GFP_KERNEL);
 755         if (!adapter->napi)
 756                 return -ENOMEM;
 757
 758         for (i = 0; i < adapter->req_rx_queues; i++) {
 759                 netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
 760                 netif_napi_add(adapter->netdev, &adapter->napi[i],
 761                                ibmvnic_poll, NAPI_POLL_WEIGHT);
 762         }
 763
 764         adapter->num_active_rx_napi = adapter->req_rx_queues;
 765         return 0;
 766 }
 767
 768 static void release_napi(struct ibmvnic_adapter *adapter)
 769 {
 770         int i;
 771
 772         if (!adapter->napi)
 773                 return;
 774
 775         for (i = 0; i < adapter->num_active_rx_napi; i++) {
 776                 if (&adapter->napi[i]) {
 777                         netdev_dbg(adapter->netdev,
 778                                    "Releasing napi[%d]\n", i);
 779                         netif_napi_del(&adapter->napi[i]);
 780                 }
 781         }
 782
 783         kfree(adapter->napi);
 784         adapter->napi = NULL;
 785         adapter->num_active_rx_napi = 0;
 786         adapter->napi_enabled = false;
 787 }
 788
 789 static int ibmvnic_login(struct net_device *netdev)
 790 {
 791         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 792         unsigned long timeout = msecs_to_jiffies(30000);
 793         int retry_count = 0;
 794         bool retry;
 795         int rc;
 796
 797         do {
 798                 retry = false;
 799                 if (retry_count > IBMVNIC_MAX_QUEUES) {
 800                         netdev_warn(netdev, "Login attempts exceeded\n");
 801                         return -1;
 802                 }
 803
 804                 adapter->init_done_rc = 0;
 805                 reinit_completion(&adapter->init_done);
 806                 rc = send_login(adapter);
 807                 if (rc) {
 808                         netdev_warn(netdev, "Unable to login\n");
 809                         return rc;
 810                 }
 811
 812                 if (!wait_for_completion_timeout(&adapter->init_done,
 813                                                  timeout)) {
 814                         netdev_warn(netdev, "Login timed out\n");
 815                         return -1;
 816                 }
 817
 818                 if (adapter->init_done_rc == PARTIALSUCCESS) {
 819                         retry_count++;
 820                         release_sub_crqs(adapter, 1);
 821
 822                         retry = true;
 823                         netdev_dbg(netdev,
 824                                    "Received partial success, retrying...\n");
 825                         adapter->init_done_rc = 0;
 826                         reinit_completion(&adapter->init_done);
 827                         send_cap_queries(adapter);
 828                         if (!wait_for_completion_timeout(&adapter->init_done,
 829                                                          timeout)) {
 830                                 netdev_warn(netdev,
 831                                             "Capabilities query timed out\n");
 832                                 return -1;
 833                         }
 834
 835                         rc = init_sub_crqs(adapter);
 836                         if (rc) {
 837                                 netdev_warn(netdev,
 838                                             "SCRQ initialization failed\n");
 839                                 return -1;
 840                         }
 841
 842                         rc = init_sub_crq_irqs(adapter);
 843                         if (rc) {
 844                                 netdev_warn(netdev,
 845                                             "SCRQ irq initialization failed\n");
 846                                 return -1;
 847                         }
 848                 } else if (adapter->init_done_rc) {
 849                         netdev_warn(netdev, "Adapter login failed\n");
 850                         return -1;
 851                 }
 852         } while (retry);
 853
 854         /* handle pending MAC address changes after successful login */
 855         if (adapter->mac_change_pending) {
 856                 __ibmvnic_set_mac(netdev, &adapter->desired.mac);
 857                 adapter->mac_change_pending = false;
 858         }
 859
 860         return 0;
 861 }
 862
 863 static void release_login_buffer(struct ibmvnic_adapter *adapter)
 864 {
 865         kfree(adapter->login_buf);
 866         adapter->login_buf = NULL;
 867 }
 868
 869 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
 870 {
 871         kfree(adapter->login_rsp_buf);
 872         adapter->login_rsp_buf = NULL;
 873 }
 874
 875 static void release_resources(struct ibmvnic_adapter *adapter)
 876 {
 877         release_vpd_data(adapter);
 878
 879         release_tx_pools(adapter);
 880         release_rx_pools(adapter);
 881
 882         release_napi(adapter);
 883         release_login_rsp_buffer(adapter);
 884 }
 885
 886 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
 887 {
 888         struct net_device *netdev = adapter->netdev;
 889         unsigned long timeout = msecs_to_jiffies(30000);
 890         union ibmvnic_crq crq;
 891         bool resend;
 892         int rc;
 893
 894         netdev_dbg(netdev, "setting link state %d\n", link_state);
 895
 896         memset(&crq, 0, sizeof(crq));
 897         crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
 898         crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
 899         crq.logical_link_state.link_state = link_state;
 900
 901         do {
 902                 resend = false;
 903
 904                 reinit_completion(&adapter->init_done);
 905                 rc = ibmvnic_send_crq(adapter, &crq);
 906                 if (rc) {
 907                         netdev_err(netdev, "Failed to set link state\n");
 908                         return rc;
 909                 }
 910
 911                 if (!wait_for_completion_timeout(&adapter->init_done,
 912                                                  timeout)) {
 913                         netdev_err(netdev, "timeout setting link state\n");
 914                         return -1;
 915                 }
 916
 917                 if (adapter->init_done_rc == 1) {
 918                         /* Partuial success, delay and re-send */
 919                         mdelay(1000);
 920                         resend = true;
 921                 } else if (adapter->init_done_rc) {
 922                         netdev_warn(netdev, "Unable to set link state, rc=%d\n",
 923                                     adapter->init_done_rc);
 924                         return adapter->init_done_rc;
 925                 }
 926         } while (resend);
 927
 928         return 0;
 929 }
 930
 931 static int set_real_num_queues(struct net_device *netdev)
 932 {
 933         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 934         int rc;
 935
 936         netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
 937                    adapter->req_tx_queues, adapter->req_rx_queues);
 938
 939         rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
 940         if (rc) {
 941                 netdev_err(netdev, "failed to set the number of tx queues\n");
 942                 return rc;
 943         }
 944
 945         rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
 946         if (rc)
 947                 netdev_err(netdev, "failed to set the number of rx queues\n");
 948
 949         return rc;
 950 }
 951
 952 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
 953 {
 954         struct device *dev = &adapter->vdev->dev;
 955         union ibmvnic_crq crq;
 956         int len = 0;
 957         int rc;
 958
 959         if (adapter->vpd->buff)
 960                 len = adapter->vpd->len;
 961
 962         init_completion(&adapter->fw_done);
 963         crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
 964         crq.get_vpd_size.cmd = GET_VPD_SIZE;
 965         rc = ibmvnic_send_crq(adapter, &crq);
 966         if (rc)
 967                 return rc;
 968         wait_for_completion(&adapter->fw_done);
 969
 970         if (!adapter->vpd->len)
 971                 return -ENODATA;
 972
 973         if (!adapter->vpd->buff)
 974                 adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
 975         else if (adapter->vpd->len != len)
 976                 adapter->vpd->buff =
 977                         krealloc(adapter->vpd->buff,
 978                                  adapter->vpd->len, GFP_KERNEL);
 979
 980         if (!adapter->vpd->buff) {
 981                 dev_err(dev, "Could allocate VPD buffer\n");
 982                 return -ENOMEM;
 983         }
 984
 985         adapter->vpd->dma_addr =
 986                 dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
 987                                DMA_FROM_DEVICE);
 988         if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
 989                 dev_err(dev, "Could not map VPD buffer\n");
 990                 kfree(adapter->vpd->buff);
 991                 adapter->vpd->buff = NULL;
 992                 return -ENOMEM;
 993         }
 994
 995         reinit_completion(&adapter->fw_done);
 996         crq.get_vpd.first = IBMVNIC_CRQ_CMD;
 997         crq.get_vpd.cmd = GET_VPD;
 998         crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
 999         crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
1000         rc = ibmvnic_send_crq(adapter, &crq);
1001         if (rc) {
1002                 kfree(adapter->vpd->buff);
1003                 adapter->vpd->buff = NULL;
1004                 return rc;
1005         }
1006         wait_for_completion(&adapter->fw_done);
1007
1008         return 0;
1009 }
1010
1011 static int init_resources(struct ibmvnic_adapter *adapter)
1012 {
1013         struct net_device *netdev = adapter->netdev;
1014         int rc;
1015
1016         rc = set_real_num_queues(netdev);
1017         if (rc)
1018                 return rc;
1019
1020         adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1021         if (!adapter->vpd)
1022                 return -ENOMEM;
1023
1024         /* Vital Product Data (VPD) */
1025         rc = ibmvnic_get_vpd(adapter);
1026         if (rc) {
1027                 netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1028                 return rc;
1029         }
1030
1031         adapter->map_id = 1;
1032
1033         rc = init_napi(adapter);
1034         if (rc)
1035                 return rc;
1036
1037         send_map_query(adapter);
1038
1039         rc = init_rx_pools(netdev);
1040         if (rc)
1041                 return rc;
1042
1043         rc = init_tx_pools(netdev);
1044         return rc;
1045 }
1046
1047 static int __ibmvnic_open(struct net_device *netdev)
1048 {
1049         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1050         enum vnic_state prev_state = adapter->state;
1051         int i, rc;
1052
1053         adapter->state = VNIC_OPENING;
1054         replenish_pools(adapter);
1055         ibmvnic_napi_enable(adapter);
1056
1057         /* We're ready to receive frames, enable the sub-crq interrupts and
1058          * set the logical link state to up
1059          */
1060         for (i = 0; i < adapter->req_rx_queues; i++) {
1061                 netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1062                 if (prev_state == VNIC_CLOSED)
1063                         enable_irq(adapter->rx_scrq[i]->irq);
1064                 enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1065         }
1066
1067         for (i = 0; i < adapter->req_tx_queues; i++) {
1068                 netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1069                 if (prev_state == VNIC_CLOSED)
1070                         enable_irq(adapter->tx_scrq[i]->irq);
1071                 enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1072         }
1073
1074         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1075         if (rc) {
1076                 for (i = 0; i < adapter->req_rx_queues; i++)
1077                         napi_disable(&adapter->napi[i]);
1078                 release_resources(adapter);
1079                 return rc;
1080         }
1081
1082         netif_tx_start_all_queues(netdev);
1083
1084         if (prev_state == VNIC_CLOSED) {
1085                 for (i = 0; i < adapter->req_rx_queues; i++)
1086                         napi_schedule(&adapter->napi[i]);
1087         }
1088
1089         adapter->state = VNIC_OPEN;
1090         return rc;
1091 }
1092
1093 static int ibmvnic_open(struct net_device *netdev)
1094 {
1095         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1096         int rc;
1097
1098         /* If device failover is pending, just set device state and return.
1099          * Device operation will be handled by reset routine.
1100          */
1101         if (adapter->failover_pending) {
1102                 adapter->state = VNIC_OPEN;
1103                 return 0;
1104         }
1105
1106         if (adapter->state != VNIC_CLOSED) {
1107                 rc = ibmvnic_login(netdev);
1108                 if (rc)
1109                         return rc;
1110
1111                 rc = init_resources(adapter);
1112                 if (rc) {
1113                         netdev_err(netdev, "failed to initialize resources\n");
1114                         release_resources(adapter);
1115                         return rc;
1116                 }
1117         }
1118
1119         rc = __ibmvnic_open(netdev);
1120         netif_carrier_on(netdev);
1121
1122         return rc;
1123 }
1124
1125 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1126 {
1127         struct ibmvnic_rx_pool *rx_pool;
1128         struct ibmvnic_rx_buff *rx_buff;
1129         u64 rx_entries;
1130         int rx_scrqs;
1131         int i, j;
1132
1133         if (!adapter->rx_pool)
1134                 return;
1135
1136         rx_scrqs = adapter->num_active_rx_pools;
1137         rx_entries = adapter->req_rx_add_entries_per_subcrq;
1138
1139         /* Free any remaining skbs in the rx buffer pools */
1140         for (i = 0; i < rx_scrqs; i++) {
1141                 rx_pool = &adapter->rx_pool[i];
1142                 if (!rx_pool || !rx_pool->rx_buff)
1143                         continue;
1144
1145                 netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1146                 for (j = 0; j < rx_entries; j++) {
1147                         rx_buff = &rx_pool->rx_buff[j];
1148                         if (rx_buff && rx_buff->skb) {
1149                                 dev_kfree_skb_any(rx_buff->skb);
1150                                 rx_buff->skb = NULL;
1151                         }
1152                 }
1153         }
1154 }
1155
1156 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1157                               struct ibmvnic_tx_pool *tx_pool)
1158 {
1159         struct ibmvnic_tx_buff *tx_buff;
1160         u64 tx_entries;
1161         int i;
1162
1163         if (!tx_pool || !tx_pool->tx_buff)
1164                 return;
1165
1166         tx_entries = tx_pool->num_buffers;
1167
1168         for (i = 0; i < tx_entries; i++) {
1169                 tx_buff = &tx_pool->tx_buff[i];
1170                 if (tx_buff && tx_buff->skb) {
1171                         dev_kfree_skb_any(tx_buff->skb);
1172                         tx_buff->skb = NULL;
1173                 }
1174         }
1175 }
1176
1177 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1178 {
1179         int tx_scrqs;
1180         int i;
1181
1182         if (!adapter->tx_pool || !adapter->tso_pool)
1183                 return;
1184
1185         tx_scrqs = adapter->num_active_tx_pools;
1186
1187         /* Free any remaining skbs in the tx buffer pools */
1188         for (i = 0; i < tx_scrqs; i++) {
1189                 netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1190                 clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1191                 clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1192         }
1193 }
1194
1195 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1196 {
1197         struct net_device *netdev = adapter->netdev;
1198         int i;
1199
1200         if (adapter->tx_scrq) {
1201                 for (i = 0; i < adapter->req_tx_queues; i++)
1202                         if (adapter->tx_scrq[i]->irq) {
1203                                 netdev_dbg(netdev,
1204                                            "Disabling tx_scrq[%d] irq\n", i);
1205                                 disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1206                                 disable_irq(adapter->tx_scrq[i]->irq);
1207                         }
1208         }
1209
1210         if (adapter->rx_scrq) {
1211                 for (i = 0; i < adapter->req_rx_queues; i++) {
1212                         if (adapter->rx_scrq[i]->irq) {
1213                                 netdev_dbg(netdev,
1214                                            "Disabling rx_scrq[%d] irq\n", i);
1215                                 disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1216                                 disable_irq(adapter->rx_scrq[i]->irq);
1217                         }
1218                 }
1219         }
1220 }
1221
1222 static void ibmvnic_cleanup(struct net_device *netdev)
1223 {
1224         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1225
1226         /* ensure that transmissions are stopped if called by do_reset */
1227         if (adapter->resetting)
1228                 netif_tx_disable(netdev);
1229         else
1230                 netif_tx_stop_all_queues(netdev);
1231
1232         ibmvnic_napi_disable(adapter);
1233         ibmvnic_disable_irqs(adapter);
1234
1235         clean_rx_pools(adapter);
1236         clean_tx_pools(adapter);
1237 }
1238
1239 static int __ibmvnic_close(struct net_device *netdev)
1240 {
1241         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1242         int rc = 0;
1243
1244         adapter->state = VNIC_CLOSING;
1245         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1246         if (rc)
1247                 return rc;
1248         adapter->state = VNIC_CLOSED;
1249         return 0;
1250 }
1251
1252 static int ibmvnic_close(struct net_device *netdev)
1253 {
1254         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1255         int rc;
1256
1257         /* If device failover is pending, just set device state and return.
1258          * Device operation will be handled by reset routine.
1259          */
1260         if (adapter->failover_pending) {
1261                 adapter->state = VNIC_CLOSED;
1262                 return 0;
1263         }
1264
1265         rc = __ibmvnic_close(netdev);
1266         ibmvnic_cleanup(netdev);
1267
1268         return rc;
1269 }
1270
1271 /**
1272  * build_hdr_data - creates L2/L3/L4 header data buffer
1273  * @hdr_field - bitfield determining needed headers
1274  * @skb - socket buffer
1275  * @hdr_len - array of header lengths
1276  * @tot_len - total length of data
1277  *
1278  * Reads hdr_field to determine which headers are needed by firmware.
1279  * Builds a buffer containing these headers.  Saves individual header
1280  * lengths and total buffer length to be used to build descriptors.
1281  */
1282 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1283                           int *hdr_len, u8 *hdr_data)
1284 {
1285         int len = 0;
1286         u8 *hdr;
1287
1288         if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1289                 hdr_len[0] = sizeof(struct vlan_ethhdr);
1290         else
1291                 hdr_len[0] = sizeof(struct ethhdr);
1292
1293         if (skb->protocol == htons(ETH_P_IP)) {
1294                 hdr_len[1] = ip_hdr(skb)->ihl * 4;
1295                 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1296                         hdr_len[2] = tcp_hdrlen(skb);
1297                 else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1298                         hdr_len[2] = sizeof(struct udphdr);
1299         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1300                 hdr_len[1] = sizeof(struct ipv6hdr);
1301                 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1302                         hdr_len[2] = tcp_hdrlen(skb);
1303                 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1304                         hdr_len[2] = sizeof(struct udphdr);
1305         } else if (skb->protocol == htons(ETH_P_ARP)) {
1306                 hdr_len[1] = arp_hdr_len(skb->dev);
1307                 hdr_len[2] = 0;
1308         }
1309
1310         memset(hdr_data, 0, 120);
1311         if ((hdr_field >> 6) & 1) {
1312                 hdr = skb_mac_header(skb);
1313                 memcpy(hdr_data, hdr, hdr_len[0]);
1314                 len += hdr_len[0];
1315         }
1316
1317         if ((hdr_field >> 5) & 1) {
1318                 hdr = skb_network_header(skb);
1319                 memcpy(hdr_data + len, hdr, hdr_len[1]);
1320                 len += hdr_len[1];
1321         }
1322
1323         if ((hdr_field >> 4) & 1) {
1324                 hdr = skb_transport_header(skb);
1325                 memcpy(hdr_data + len, hdr, hdr_len[2]);
1326                 len += hdr_len[2];
1327         }
1328         return len;
1329 }
1330
1331 /**
1332  * create_hdr_descs - create header and header extension descriptors
1333  * @hdr_field - bitfield determining needed headers
1334  * @data - buffer containing header data
1335  * @len - length of data buffer
1336  * @hdr_len - array of individual header lengths
1337  * @scrq_arr - descriptor array
1338  *
1339  * Creates header and, if needed, header extension descriptors and
1340  * places them in a descriptor array, scrq_arr
1341  */
1342
1343 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1344                             union sub_crq *scrq_arr)
1345 {
1346         union sub_crq hdr_desc;
1347         int tmp_len = len;
1348         int num_descs = 0;
1349         u8 *data, *cur;
1350         int tmp;
1351
1352         while (tmp_len > 0) {
1353                 cur = hdr_data + len - tmp_len;
1354
1355                 memset(&hdr_desc, 0, sizeof(hdr_desc));
1356                 if (cur != hdr_data) {
1357                         data = hdr_desc.hdr_ext.data;
1358                         tmp = tmp_len > 29 ? 29 : tmp_len;
1359                         hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1360                         hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1361                         hdr_desc.hdr_ext.len = tmp;
1362                 } else {
1363                         data = hdr_desc.hdr.data;
1364                         tmp = tmp_len > 24 ? 24 : tmp_len;
1365                         hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1366                         hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1367                         hdr_desc.hdr.len = tmp;
1368                         hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1369                         hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1370                         hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1371                         hdr_desc.hdr.flag = hdr_field << 1;
1372                 }
1373                 memcpy(data, cur, tmp);
1374                 tmp_len -= tmp;
1375                 *scrq_arr = hdr_desc;
1376                 scrq_arr++;
1377                 num_descs++;
1378         }
1379
1380         return num_descs;
1381 }
1382
1383 /**
1384  * build_hdr_descs_arr - build a header descriptor array
1385  * @skb - socket buffer
1386  * @num_entries - number of descriptors to be sent
1387  * @subcrq - first TX descriptor
1388  * @hdr_field - bit field determining which headers will be sent
1389  *
1390  * This function will build a TX descriptor array with applicable
1391  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1392  */
1393
1394 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1395                                 int *num_entries, u8 hdr_field)
1396 {
1397         int hdr_len[3] = {0, 0, 0};
1398         int tot_len;
1399         u8 *hdr_data = txbuff->hdr_data;
1400
1401         tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1402                                  txbuff->hdr_data);
1403         *num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1404                          txbuff->indir_arr + 1);
1405 }
1406
1407 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1408                                     struct net_device *netdev)
1409 {
1410         /* For some backing devices, mishandling of small packets
1411          * can result in a loss of connection or TX stall. Device
1412          * architects recommend that no packet should be smaller
1413          * than the minimum MTU value provided to the driver, so
1414          * pad any packets to that length
1415          */
1416         if (skb->len < netdev->min_mtu)
1417                 return skb_put_padto(skb, netdev->min_mtu);
1418
1419         return 0;
1420 }
1421
1422 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1423 {
1424         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1425         int queue_num = skb_get_queue_mapping(skb);
1426         u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1427         struct device *dev = &adapter->vdev->dev;
1428         struct ibmvnic_tx_buff *tx_buff = NULL;
1429         struct ibmvnic_sub_crq_queue *tx_scrq;
1430         struct ibmvnic_tx_pool *tx_pool;
1431         unsigned int tx_send_failed = 0;
1432         unsigned int tx_map_failed = 0;
1433         unsigned int tx_dropped = 0;
1434         unsigned int tx_packets = 0;
1435         unsigned int tx_bytes = 0;
1436         dma_addr_t data_dma_addr;
1437         struct netdev_queue *txq;
1438         unsigned long lpar_rc;
1439         union sub_crq tx_crq;
1440         unsigned int offset;
1441         int num_entries = 1;
1442         unsigned char *dst;
1443         u64 *handle_array;
1444         int index = 0;
1445         u8 proto = 0;
1446         netdev_tx_t ret = NETDEV_TX_OK;
1447
1448         if (adapter->resetting) {
1449                 if (!netif_subqueue_stopped(netdev, skb))
1450                         netif_stop_subqueue(netdev, queue_num);
1451                 dev_kfree_skb_any(skb);
1452
1453                 tx_send_failed++;
1454                 tx_dropped++;
1455                 ret = NETDEV_TX_OK;
1456                 goto out;
1457         }
1458
1459         if (ibmvnic_xmit_workarounds(skb, netdev)) {
1460                 tx_dropped++;
1461                 tx_send_failed++;
1462                 ret = NETDEV_TX_OK;
1463                 goto out;
1464         }
1465         if (skb_is_gso(skb))
1466                 tx_pool = &adapter->tso_pool[queue_num];
1467         else
1468                 tx_pool = &adapter->tx_pool[queue_num];
1469
1470         tx_scrq = adapter->tx_scrq[queue_num];
1471         txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1472         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1473                 be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1474
1475         index = tx_pool->free_map[tx_pool->consumer_index];
1476
1477         if (index == IBMVNIC_INVALID_MAP) {
1478                 dev_kfree_skb_any(skb);
1479                 tx_send_failed++;
1480                 tx_dropped++;
1481                 ret = NETDEV_TX_OK;
1482                 goto out;
1483         }
1484
1485         tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1486
1487         offset = index * tx_pool->buf_size;
1488         dst = tx_pool->long_term_buff.buff + offset;
1489         memset(dst, 0, tx_pool->buf_size);
1490         data_dma_addr = tx_pool->long_term_buff.addr + offset;
1491
1492         if (skb_shinfo(skb)->nr_frags) {
1493                 int cur, i;
1494
1495                 /* Copy the head */
1496                 skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1497                 cur = skb_headlen(skb);
1498
1499                 /* Copy the frags */
1500                 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1501                         const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1502
1503                         memcpy(dst + cur,
1504                                page_address(skb_frag_page(frag)) +
1505                                frag->page_offset, skb_frag_size(frag));
1506                         cur += skb_frag_size(frag);
1507                 }
1508         } else {
1509                 skb_copy_from_linear_data(skb, dst, skb->len);
1510         }
1511
1512         tx_pool->consumer_index =
1513             (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1514
1515         tx_buff = &tx_pool->tx_buff[index];
1516         tx_buff->skb = skb;
1517         tx_buff->data_dma[0] = data_dma_addr;
1518         tx_buff->data_len[0] = skb->len;
1519         tx_buff->index = index;
1520         tx_buff->pool_index = queue_num;
1521         tx_buff->last_frag = true;
1522
1523         memset(&tx_crq, 0, sizeof(tx_crq));
1524         tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1525         tx_crq.v1.type = IBMVNIC_TX_DESC;
1526         tx_crq.v1.n_crq_elem = 1;
1527         tx_crq.v1.n_sge = 1;
1528         tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1529
1530         if (skb_is_gso(skb))
1531                 tx_crq.v1.correlator =
1532                         cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1533         else
1534                 tx_crq.v1.correlator = cpu_to_be32(index);
1535         tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1536         tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1537         tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1538
1539         if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1540                 tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1541                 tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1542         }
1543
1544         if (skb->protocol == htons(ETH_P_IP)) {
1545                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1546                 proto = ip_hdr(skb)->protocol;
1547         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1548                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1549                 proto = ipv6_hdr(skb)->nexthdr;
1550         }
1551
1552         if (proto == IPPROTO_TCP)
1553                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1554         else if (proto == IPPROTO_UDP)
1555                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1556
1557         if (skb->ip_summed == CHECKSUM_PARTIAL) {
1558                 tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1559                 hdrs += 2;
1560         }
1561         if (skb_is_gso(skb)) {
1562                 tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1563                 tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1564                 hdrs += 2;
1565         }
1566         /* determine if l2/3/4 headers are sent to firmware */
1567         if ((*hdrs >> 7) & 1) {
1568                 build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1569                 tx_crq.v1.n_crq_elem = num_entries;
1570                 tx_buff->num_entries = num_entries;
1571                 tx_buff->indir_arr[0] = tx_crq;
1572                 tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1573                                                     sizeof(tx_buff->indir_arr),
1574                                                     DMA_TO_DEVICE);
1575                 if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1576                         dev_kfree_skb_any(skb);
1577                         tx_buff->skb = NULL;
1578                         if (!firmware_has_feature(FW_FEATURE_CMO))
1579                                 dev_err(dev, "tx: unable to map descriptor array\n");
1580                         tx_map_failed++;
1581                         tx_dropped++;
1582                         ret = NETDEV_TX_OK;
1583                         goto tx_err_out;
1584                 }
1585                 lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1586                                                (u64)tx_buff->indir_dma,
1587                                                (u64)num_entries);
1588         } else {
1589                 tx_buff->num_entries = num_entries;
1590                 lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1591                                       &tx_crq);
1592         }
1593         if (lpar_rc != H_SUCCESS) {
1594                 if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1595                         dev_err_ratelimited(dev, "tx: send failed\n");
1596                 dev_kfree_skb_any(skb);
1597                 tx_buff->skb = NULL;
1598
1599                 if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1600                         /* Disable TX and report carrier off if queue is closed
1601                          * or pending failover.
1602                          * Firmware guarantees that a signal will be sent to the
1603                          * driver, triggering a reset or some other action.
1604                          */
1605                         netif_tx_stop_all_queues(netdev);
1606                         netif_carrier_off(netdev);
1607                 }
1608
1609                 tx_send_failed++;
1610                 tx_dropped++;
1611                 ret = NETDEV_TX_OK;
1612                 goto tx_err_out;
1613         }
1614
1615         if (atomic_add_return(num_entries, &tx_scrq->used)
1616                                         >= adapter->req_tx_entries_per_subcrq) {
1617                 netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1618                 netif_stop_subqueue(netdev, queue_num);
1619         }
1620
1621         tx_packets++;
1622         tx_bytes += skb->len;
1623         txq->trans_start = jiffies;
1624         ret = NETDEV_TX_OK;
1625         goto out;
1626
1627 tx_err_out:
1628         /* roll back consumer index and map array*/
1629         if (tx_pool->consumer_index == 0)
1630                 tx_pool->consumer_index =
1631                         tx_pool->num_buffers - 1;
1632         else
1633                 tx_pool->consumer_index--;
1634         tx_pool->free_map[tx_pool->consumer_index] = index;
1635 out:
1636         netdev->stats.tx_dropped += tx_dropped;
1637         netdev->stats.tx_bytes += tx_bytes;
1638         netdev->stats.tx_packets += tx_packets;
1639         adapter->tx_send_failed += tx_send_failed;
1640         adapter->tx_map_failed += tx_map_failed;
1641         adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1642         adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1643         adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1644
1645         return ret;
1646 }
1647
1648 static void ibmvnic_set_multi(struct net_device *netdev)
1649 {
1650         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1651         struct netdev_hw_addr *ha;
1652         union ibmvnic_crq crq;
1653
1654         memset(&crq, 0, sizeof(crq));
1655         crq.request_capability.first = IBMVNIC_CRQ_CMD;
1656         crq.request_capability.cmd = REQUEST_CAPABILITY;
1657
1658         if (netdev->flags & IFF_PROMISC) {
1659                 if (!adapter->promisc_supported)
1660                         return;
1661         } else {
1662                 if (netdev->flags & IFF_ALLMULTI) {
1663                         /* Accept all multicast */
1664                         memset(&crq, 0, sizeof(crq));
1665                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1666                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1667                         crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1668                         ibmvnic_send_crq(adapter, &crq);
1669                 } else if (netdev_mc_empty(netdev)) {
1670                         /* Reject all multicast */
1671                         memset(&crq, 0, sizeof(crq));
1672                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1673                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1674                         crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1675                         ibmvnic_send_crq(adapter, &crq);
1676                 } else {
1677                         /* Accept one or more multicast(s) */
1678                         netdev_for_each_mc_addr(ha, netdev) {
1679                                 memset(&crq, 0, sizeof(crq));
1680                                 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1681                                 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1682                                 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1683                                 ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1684                                                 ha->addr);
1685                                 ibmvnic_send_crq(adapter, &crq);
1686                         }
1687                 }
1688         }
1689 }
1690
1691 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p)
1692 {
1693         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1694         struct sockaddr *addr = p;
1695         union ibmvnic_crq crq;
1696         int rc;
1697
1698         if (!is_valid_ether_addr(addr->sa_data))
1699                 return -EADDRNOTAVAIL;
1700
1701         memset(&crq, 0, sizeof(crq));
1702         crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1703         crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1704         ether_addr_copy(&crq.change_mac_addr.mac_addr[0], addr->sa_data);
1705
1706         init_completion(&adapter->fw_done);
1707         rc = ibmvnic_send_crq(adapter, &crq);
1708         if (rc)
1709                 return rc;
1710         wait_for_completion(&adapter->fw_done);
1711         /* netdev->dev_addr is changed in handle_change_mac_rsp function */
1712         return adapter->fw_done_rc ? -EIO : 0;
1713 }
1714
1715 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1716 {
1717         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1718         struct sockaddr *addr = p;
1719         int rc;
1720
1721         if (adapter->state == VNIC_PROBED) {
1722                 memcpy(&adapter->desired.mac, addr, sizeof(struct sockaddr));
1723                 adapter->mac_change_pending = true;
1724                 return 0;
1725         }
1726
1727         rc = __ibmvnic_set_mac(netdev, addr);
1728
1729         return rc;
1730 }
1731
1732 /**
1733  * do_reset returns zero if we are able to keep processing reset events, or
1734  * non-zero if we hit a fatal error and must halt.
1735  */
1736 static int do_reset(struct ibmvnic_adapter *adapter,
1737                     struct ibmvnic_rwi *rwi, u32 reset_state)
1738 {
1739         u64 old_num_rx_queues, old_num_tx_queues;
1740         struct net_device *netdev = adapter->netdev;
1741         int i, rc;
1742
1743         netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1744                    rwi->reset_reason);
1745
1746         netif_carrier_off(netdev);
1747         adapter->reset_reason = rwi->reset_reason;
1748
1749         old_num_rx_queues = adapter->req_rx_queues;
1750         old_num_tx_queues = adapter->req_tx_queues;
1751
1752         ibmvnic_cleanup(netdev);
1753
1754         if (adapter->reset_reason != VNIC_RESET_MOBILITY &&
1755             adapter->reset_reason != VNIC_RESET_FAILOVER) {
1756                 rc = __ibmvnic_close(netdev);
1757                 if (rc)
1758                         return rc;
1759         }
1760
1761         if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1762             adapter->wait_for_reset) {
1763                 release_resources(adapter);
1764                 release_sub_crqs(adapter, 1);
1765                 release_crq_queue(adapter);
1766         }
1767
1768         if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1769                 /* remove the closed state so when we call open it appears
1770                  * we are coming from the probed state.
1771                  */
1772                 adapter->state = VNIC_PROBED;
1773
1774                 if (adapter->wait_for_reset) {
1775                         rc = init_crq_queue(adapter);
1776                 } else if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1777                         rc = ibmvnic_reenable_crq_queue(adapter);
1778                         release_sub_crqs(adapter, 1);
1779                 } else {
1780                         rc = ibmvnic_reset_crq(adapter);
1781                         if (!rc)
1782                                 rc = vio_enable_interrupts(adapter->vdev);
1783                 }
1784
1785                 if (rc) {
1786                         netdev_err(adapter->netdev,
1787                                    "Couldn't initialize crq. rc=%d\n", rc);
1788                         return rc;
1789                 }
1790
1791                 rc = ibmvnic_reset_init(adapter);
1792                 if (rc)
1793                         return IBMVNIC_INIT_FAILED;
1794
1795                 /* If the adapter was in PROBE state prior to the reset,
1796                  * exit here.
1797                  */
1798                 if (reset_state == VNIC_PROBED)
1799                         return 0;
1800
1801                 rc = ibmvnic_login(netdev);
1802                 if (rc) {
1803                         adapter->state = reset_state;
1804                         return rc;
1805                 }
1806
1807                 if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1808                     adapter->wait_for_reset) {
1809                         rc = init_resources(adapter);
1810                         if (rc)
1811                                 return rc;
1812                 } else if (adapter->req_rx_queues != old_num_rx_queues ||
1813                            adapter->req_tx_queues != old_num_tx_queues) {
1814                         release_rx_pools(adapter);
1815                         release_tx_pools(adapter);
1816                         release_napi(adapter);
1817                         release_vpd_data(adapter);
1818
1819                         rc = init_resources(adapter);
1820                         if (rc)
1821                                 return rc;
1822
1823                 } else {
1824                         rc = reset_tx_pools(adapter);
1825                         if (rc)
1826                                 return rc;
1827
1828                         rc = reset_rx_pools(adapter);
1829                         if (rc)
1830                                 return rc;
1831                 }
1832                 ibmvnic_disable_irqs(adapter);
1833         }
1834         adapter->state = VNIC_CLOSED;
1835
1836         if (reset_state == VNIC_CLOSED)
1837                 return 0;
1838
1839         rc = __ibmvnic_open(netdev);
1840         if (rc) {
1841                 if (list_empty(&adapter->rwi_list))
1842                         adapter->state = VNIC_CLOSED;
1843                 else
1844                         adapter->state = reset_state;
1845
1846                 return 0;
1847         }
1848
1849         /* kick napi */
1850         for (i = 0; i < adapter->req_rx_queues; i++)
1851                 napi_schedule(&adapter->napi[i]);
1852
1853         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
1854             adapter->reset_reason != VNIC_RESET_CHANGE_PARAM)
1855                 netdev_notify_peers(netdev);
1856
1857         netif_carrier_on(netdev);
1858
1859         return 0;
1860 }
1861
1862 static int do_hard_reset(struct ibmvnic_adapter *adapter,
1863                          struct ibmvnic_rwi *rwi, u32 reset_state)
1864 {
1865         struct net_device *netdev = adapter->netdev;
1866         int rc;
1867
1868         netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
1869                    rwi->reset_reason);
1870
1871         netif_carrier_off(netdev);
1872         adapter->reset_reason = rwi->reset_reason;
1873
1874         ibmvnic_cleanup(netdev);
1875         release_resources(adapter);
1876         release_sub_crqs(adapter, 0);
1877         release_crq_queue(adapter);
1878
1879         /* remove the closed state so when we call open it appears
1880          * we are coming from the probed state.
1881          */
1882         adapter->state = VNIC_PROBED;
1883
1884         rc = init_crq_queue(adapter);
1885         if (rc) {
1886                 netdev_err(adapter->netdev,
1887                            "Couldn't initialize crq. rc=%d\n", rc);
1888                 return rc;
1889         }
1890
1891         rc = ibmvnic_init(adapter);
1892         if (rc)
1893                 return rc;
1894
1895         /* If the adapter was in PROBE state prior to the reset,
1896          * exit here.
1897          */
1898         if (reset_state == VNIC_PROBED)
1899                 return 0;
1900
1901         rc = ibmvnic_login(netdev);
1902         if (rc) {
1903                 adapter->state = VNIC_PROBED;
1904                 return 0;
1905         }
1906
1907         rc = init_resources(adapter);
1908         if (rc)
1909                 return rc;
1910
1911         ibmvnic_disable_irqs(adapter);
1912         adapter->state = VNIC_CLOSED;
1913
1914         if (reset_state == VNIC_CLOSED)
1915                 return 0;
1916
1917         rc = __ibmvnic_open(netdev);
1918         if (rc) {
1919                 if (list_empty(&adapter->rwi_list))
1920                         adapter->state = VNIC_CLOSED;
1921                 else
1922                         adapter->state = reset_state;
1923
1924                 return 0;
1925         }
1926
1927         netif_carrier_on(netdev);
1928
1929         return 0;
1930 }
1931
1932 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1933 {
1934         struct ibmvnic_rwi *rwi;
1935
1936         mutex_lock(&adapter->rwi_lock);
1937
1938         if (!list_empty(&adapter->rwi_list)) {
1939                 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1940                                        list);
1941                 list_del(&rwi->list);
1942         } else {
1943                 rwi = NULL;
1944         }
1945
1946         mutex_unlock(&adapter->rwi_lock);
1947         return rwi;
1948 }
1949
1950 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1951 {
1952         struct ibmvnic_rwi *rwi;
1953
1954         rwi = get_next_rwi(adapter);
1955         while (rwi) {
1956                 kfree(rwi);
1957                 rwi = get_next_rwi(adapter);
1958         }
1959 }
1960
1961 static void __ibmvnic_reset(struct work_struct *work)
1962 {
1963         struct ibmvnic_rwi *rwi;
1964         struct ibmvnic_adapter *adapter;
1965         struct net_device *netdev;
1966         bool we_lock_rtnl = false;
1967         u32 reset_state;
1968         int rc = 0;
1969
1970         adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1971         netdev = adapter->netdev;
1972
1973         /* netif_set_real_num_xx_queues needs to take rtnl lock here
1974          * unless wait_for_reset is set, in which case the rtnl lock
1975          * has already been taken before initializing the reset
1976          */
1977         if (!adapter->wait_for_reset) {
1978                 rtnl_lock();
1979                 we_lock_rtnl = true;
1980         }
1981         reset_state = adapter->state;
1982
1983         rwi = get_next_rwi(adapter);
1984         while (rwi) {
1985                 if (adapter->force_reset_recovery) {
1986                         adapter->force_reset_recovery = false;
1987                         rc = do_hard_reset(adapter, rwi, reset_state);
1988                 } else {
1989                         rc = do_reset(adapter, rwi, reset_state);
1990                 }
1991                 kfree(rwi);
1992                 if (rc && rc != IBMVNIC_INIT_FAILED &&
1993                     !adapter->force_reset_recovery)
1994                         break;
1995
1996                 rwi = get_next_rwi(adapter);
1997         }
1998
1999         if (adapter->wait_for_reset) {
2000                 adapter->wait_for_reset = false;
2001                 adapter->reset_done_rc = rc;
2002                 complete(&adapter->reset_done);
2003         }
2004
2005         if (rc) {
2006                 netdev_dbg(adapter->netdev, "Reset failed\n");
2007                 free_all_rwi(adapter);
2008         }
2009
2010         adapter->resetting = false;
2011         if (we_lock_rtnl)
2012                 rtnl_unlock();
2013 }
2014
2015 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2016                          enum ibmvnic_reset_reason reason)
2017 {
2018         struct list_head *entry, *tmp_entry;
2019         struct ibmvnic_rwi *rwi, *tmp;
2020         struct net_device *netdev = adapter->netdev;
2021         int ret;
2022
2023         if (adapter->state == VNIC_REMOVING ||
2024             adapter->state == VNIC_REMOVED ||
2025             adapter->failover_pending) {
2026                 ret = EBUSY;
2027                 netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2028                 goto err;
2029         }
2030
2031         if (adapter->state == VNIC_PROBING) {
2032                 netdev_warn(netdev, "Adapter reset during probe\n");
2033                 ret = adapter->init_done_rc = EAGAIN;
2034                 goto err;
2035         }
2036
2037         mutex_lock(&adapter->rwi_lock);
2038
2039         list_for_each(entry, &adapter->rwi_list) {
2040                 tmp = list_entry(entry, struct ibmvnic_rwi, list);
2041                 if (tmp->reset_reason == reason) {
2042                         netdev_dbg(netdev, "Skipping matching reset\n");
2043                         mutex_unlock(&adapter->rwi_lock);
2044                         ret = EBUSY;
2045                         goto err;
2046                 }
2047         }
2048
2049         rwi = kzalloc(sizeof(*rwi), GFP_KERNEL);
2050         if (!rwi) {
2051                 mutex_unlock(&adapter->rwi_lock);
2052                 ibmvnic_close(netdev);
2053                 ret = ENOMEM;
2054                 goto err;
2055         }
2056         /* if we just received a transport event,
2057          * flush reset queue and process this reset
2058          */
2059         if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2060                 list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2061                         list_del(entry);
2062         }
2063         rwi->reset_reason = reason;
2064         list_add_tail(&rwi->list, &adapter->rwi_list);
2065         mutex_unlock(&adapter->rwi_lock);
2066         adapter->resetting = true;
2067         netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2068         schedule_work(&adapter->ibmvnic_reset);
2069
2070         return 0;
2071 err:
2072         if (adapter->wait_for_reset)
2073                 adapter->wait_for_reset = false;
2074         return -ret;
2075 }
2076
2077 static void ibmvnic_tx_timeout(struct net_device *dev)
2078 {
2079         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2080
2081         ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2082 }
2083
2084 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2085                                   struct ibmvnic_rx_buff *rx_buff)
2086 {
2087         struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2088
2089         rx_buff->skb = NULL;
2090
2091         pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2092         pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2093
2094         atomic_dec(&pool->available);
2095 }
2096
2097 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2098 {
2099         struct net_device *netdev = napi->dev;
2100         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2101         int scrq_num = (int)(napi - adapter->napi);
2102         int frames_processed = 0;
2103
2104 restart_poll:
2105         while (frames_processed < budget) {
2106                 struct sk_buff *skb;
2107                 struct ibmvnic_rx_buff *rx_buff;
2108                 union sub_crq *next;
2109                 u32 length;
2110                 u16 offset;
2111                 u8 flags = 0;
2112
2113                 if (unlikely(adapter->resetting &&
2114                              adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2115                         enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2116                         napi_complete_done(napi, frames_processed);
2117                         return frames_processed;
2118                 }
2119
2120                 if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2121                         break;
2122                 next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2123                 rx_buff =
2124                     (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2125                                                           rx_comp.correlator);
2126                 /* do error checking */
2127                 if (next->rx_comp.rc) {
2128                         netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2129                                    be16_to_cpu(next->rx_comp.rc));
2130                         /* free the entry */
2131                         next->rx_comp.first = 0;
2132                         dev_kfree_skb_any(rx_buff->skb);
2133                         remove_buff_from_pool(adapter, rx_buff);
2134                         continue;
2135                 } else if (!rx_buff->skb) {
2136                         /* free the entry */
2137                         next->rx_comp.first = 0;
2138                         remove_buff_from_pool(adapter, rx_buff);
2139                         continue;
2140                 }
2141
2142                 length = be32_to_cpu(next->rx_comp.len);
2143                 offset = be16_to_cpu(next->rx_comp.off_frame_data);
2144                 flags = next->rx_comp.flags;
2145                 skb = rx_buff->skb;
2146                 skb_copy_to_linear_data(skb, rx_buff->data + offset,
2147                                         length);
2148
2149                 /* VLAN Header has been stripped by the system firmware and
2150                  * needs to be inserted by the driver
2151                  */
2152                 if (adapter->rx_vlan_header_insertion &&
2153                     (flags & IBMVNIC_VLAN_STRIPPED))
2154                         __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2155                                                ntohs(next->rx_comp.vlan_tci));
2156
2157                 /* free the entry */
2158                 next->rx_comp.first = 0;
2159                 remove_buff_from_pool(adapter, rx_buff);
2160
2161                 skb_put(skb, length);
2162                 skb->protocol = eth_type_trans(skb, netdev);
2163                 skb_record_rx_queue(skb, scrq_num);
2164
2165                 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2166                     flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2167                         skb->ip_summed = CHECKSUM_UNNECESSARY;
2168                 }
2169
2170                 length = skb->len;
2171                 napi_gro_receive(napi, skb); /* send it up */
2172                 netdev->stats.rx_packets++;
2173                 netdev->stats.rx_bytes += length;
2174                 adapter->rx_stats_buffers[scrq_num].packets++;
2175                 adapter->rx_stats_buffers[scrq_num].bytes += length;
2176                 frames_processed++;
2177         }
2178
2179         if (adapter->state != VNIC_CLOSING)
2180                 replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2181
2182         if (frames_processed < budget) {
2183                 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2184                 napi_complete_done(napi, frames_processed);
2185                 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2186                     napi_reschedule(napi)) {
2187                         disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2188                         goto restart_poll;
2189                 }
2190         }
2191         return frames_processed;
2192 }
2193
2194 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2195 {
2196         int rc, ret;
2197
2198         adapter->fallback.mtu = adapter->req_mtu;
2199         adapter->fallback.rx_queues = adapter->req_rx_queues;
2200         adapter->fallback.tx_queues = adapter->req_tx_queues;
2201         adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2202         adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2203
2204         init_completion(&adapter->reset_done);
2205         adapter->wait_for_reset = true;
2206         rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2207         if (rc)
2208                 return rc;
2209         wait_for_completion(&adapter->reset_done);
2210
2211         ret = 0;
2212         if (adapter->reset_done_rc) {
2213                 ret = -EIO;
2214                 adapter->desired.mtu = adapter->fallback.mtu;
2215                 adapter->desired.rx_queues = adapter->fallback.rx_queues;
2216                 adapter->desired.tx_queues = adapter->fallback.tx_queues;
2217                 adapter->desired.rx_entries = adapter->fallback.rx_entries;
2218                 adapter->desired.tx_entries = adapter->fallback.tx_entries;
2219
2220                 init_completion(&adapter->reset_done);
2221                 adapter->wait_for_reset = true;
2222                 rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2223                 if (rc)
2224                         return ret;
2225                 wait_for_completion(&adapter->reset_done);
2226         }
2227         adapter->wait_for_reset = false;
2228
2229         return ret;
2230 }
2231
2232 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2233 {
2234         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2235
2236         adapter->desired.mtu = new_mtu + ETH_HLEN;
2237
2238         return wait_for_reset(adapter);
2239 }
2240
2241 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2242                                                 struct net_device *dev,
2243                                                 netdev_features_t features)
2244 {
2245         /* Some backing hardware adapters can not
2246          * handle packets with a MSS less than 224
2247          * or with only one segment.
2248          */
2249         if (skb_is_gso(skb)) {
2250                 if (skb_shinfo(skb)->gso_size < 224 ||
2251                     skb_shinfo(skb)->gso_segs == 1)
2252                         features &= ~NETIF_F_GSO_MASK;
2253         }
2254
2255         return features;
2256 }
2257
2258 static const struct net_device_ops ibmvnic_netdev_ops = {
2259         .ndo_open               = ibmvnic_open,
2260         .ndo_stop               = ibmvnic_close,
2261         .ndo_start_xmit         = ibmvnic_xmit,
2262         .ndo_set_rx_mode        = ibmvnic_set_multi,
2263         .ndo_set_mac_address    = ibmvnic_set_mac,
2264         .ndo_validate_addr      = eth_validate_addr,
2265         .ndo_tx_timeout         = ibmvnic_tx_timeout,
2266         .ndo_change_mtu         = ibmvnic_change_mtu,
2267         .ndo_features_check     = ibmvnic_features_check,
2268 };
2269
2270 /* ethtool functions */
2271
2272 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2273                                       struct ethtool_link_ksettings *cmd)
2274 {
2275         u32 supported, advertising;
2276
2277         supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg |
2278                           SUPPORTED_FIBRE);
2279         advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg |
2280                             ADVERTISED_FIBRE);
2281         cmd->base.speed = SPEED_1000;
2282         cmd->base.duplex = DUPLEX_FULL;
2283         cmd->base.port = PORT_FIBRE;
2284         cmd->base.phy_address = 0;
2285         cmd->base.autoneg = AUTONEG_ENABLE;
2286
2287         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
2288                                                 supported);
2289         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
2290                                                 advertising);
2291
2292         return 0;
2293 }
2294
2295 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2296                                 struct ethtool_drvinfo *info)
2297 {
2298         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2299
2300         strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2301         strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2302         strlcpy(info->fw_version, adapter->fw_version,
2303                 sizeof(info->fw_version));
2304 }
2305
2306 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2307 {
2308         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2309
2310         return adapter->msg_enable;
2311 }
2312
2313 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2314 {
2315         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2316
2317         adapter->msg_enable = data;
2318 }
2319
2320 static u32 ibmvnic_get_link(struct net_device *netdev)
2321 {
2322         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2323
2324         /* Don't need to send a query because we request a logical link up at
2325          * init and then we wait for link state indications
2326          */
2327         return adapter->logical_link_state;
2328 }
2329
2330 static void ibmvnic_get_ringparam(struct net_device *netdev,
2331                                   struct ethtool_ringparam *ring)
2332 {
2333         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2334
2335         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2336                 ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2337                 ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2338         } else {
2339                 ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2340                 ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2341         }
2342         ring->rx_mini_max_pending = 0;
2343         ring->rx_jumbo_max_pending = 0;
2344         ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2345         ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2346         ring->rx_mini_pending = 0;
2347         ring->rx_jumbo_pending = 0;
2348 }
2349
2350 static int ibmvnic_set_ringparam(struct net_device *netdev,
2351                                  struct ethtool_ringparam *ring)
2352 {
2353         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2354         int ret;
2355
2356         ret = 0;
2357         adapter->desired.rx_entries = ring->rx_pending;
2358         adapter->desired.tx_entries = ring->tx_pending;
2359
2360         ret = wait_for_reset(adapter);
2361
2362         if (!ret &&
2363             (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2364              adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2365                 netdev_info(netdev,
2366                             "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2367                             ring->rx_pending, ring->tx_pending,
2368                             adapter->req_rx_add_entries_per_subcrq,
2369                             adapter->req_tx_entries_per_subcrq);
2370         return ret;
2371 }
2372
2373 static void ibmvnic_get_channels(struct net_device *netdev,
2374                                  struct ethtool_channels *channels)
2375 {
2376         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2377
2378         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2379                 channels->max_rx = adapter->max_rx_queues;
2380                 channels->max_tx = adapter->max_tx_queues;
2381         } else {
2382                 channels->max_rx = IBMVNIC_MAX_QUEUES;
2383                 channels->max_tx = IBMVNIC_MAX_QUEUES;
2384         }
2385
2386         channels->max_other = 0;
2387         channels->max_combined = 0;
2388         channels->rx_count = adapter->req_rx_queues;
2389         channels->tx_count = adapter->req_tx_queues;
2390         channels->other_count = 0;
2391         channels->combined_count = 0;
2392 }
2393
2394 static int ibmvnic_set_channels(struct net_device *netdev,
2395                                 struct ethtool_channels *channels)
2396 {
2397         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2398         int ret;
2399
2400         ret = 0;
2401         adapter->desired.rx_queues = channels->rx_count;
2402         adapter->desired.tx_queues = channels->tx_count;
2403
2404         ret = wait_for_reset(adapter);
2405
2406         if (!ret &&
2407             (adapter->req_rx_queues != channels->rx_count ||
2408              adapter->req_tx_queues != channels->tx_count))
2409                 netdev_info(netdev,
2410                             "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2411                             channels->rx_count, channels->tx_count,
2412                             adapter->req_rx_queues, adapter->req_tx_queues);
2413         return ret;
2414
2415 }
2416
2417 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2418 {
2419         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2420         int i;
2421
2422         switch (stringset) {
2423         case ETH_SS_STATS:
2424                 for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2425                                 i++, data += ETH_GSTRING_LEN)
2426                         memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2427
2428                 for (i = 0; i < adapter->req_tx_queues; i++) {
2429                         snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2430                         data += ETH_GSTRING_LEN;
2431
2432                         snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2433                         data += ETH_GSTRING_LEN;
2434
2435                         snprintf(data, ETH_GSTRING_LEN,
2436                                  "tx%d_dropped_packets", i);
2437                         data += ETH_GSTRING_LEN;
2438                 }
2439
2440                 for (i = 0; i < adapter->req_rx_queues; i++) {
2441                         snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2442                         data += ETH_GSTRING_LEN;
2443
2444                         snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2445                         data += ETH_GSTRING_LEN;
2446
2447                         snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2448                         data += ETH_GSTRING_LEN;
2449                 }
2450                 break;
2451
2452         case ETH_SS_PRIV_FLAGS:
2453                 for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2454                         strcpy(data + i * ETH_GSTRING_LEN,
2455                                ibmvnic_priv_flags[i]);
2456                 break;
2457         default:
2458                 return;
2459         }
2460 }
2461
2462 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2463 {
2464         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2465
2466         switch (sset) {
2467         case ETH_SS_STATS:
2468                 return ARRAY_SIZE(ibmvnic_stats) +
2469                        adapter->req_tx_queues * NUM_TX_STATS +
2470                        adapter->req_rx_queues * NUM_RX_STATS;
2471         case ETH_SS_PRIV_FLAGS:
2472                 return ARRAY_SIZE(ibmvnic_priv_flags);
2473         default:
2474                 return -EOPNOTSUPP;
2475         }
2476 }
2477
2478 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2479                                       struct ethtool_stats *stats, u64 *data)
2480 {
2481         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2482         union ibmvnic_crq crq;
2483         int i, j;
2484         int rc;
2485
2486         memset(&crq, 0, sizeof(crq));
2487         crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2488         crq.request_statistics.cmd = REQUEST_STATISTICS;
2489         crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2490         crq.request_statistics.len =
2491             cpu_to_be32(sizeof(struct ibmvnic_statistics));
2492
2493         /* Wait for data to be written */
2494         init_completion(&adapter->stats_done);
2495         rc = ibmvnic_send_crq(adapter, &crq);
2496         if (rc)
2497                 return;
2498         wait_for_completion(&adapter->stats_done);
2499
2500         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2501                 data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2502                                                 ibmvnic_stats[i].offset));
2503
2504         for (j = 0; j < adapter->req_tx_queues; j++) {
2505                 data[i] = adapter->tx_stats_buffers[j].packets;
2506                 i++;
2507                 data[i] = adapter->tx_stats_buffers[j].bytes;
2508                 i++;
2509                 data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2510                 i++;
2511         }
2512
2513         for (j = 0; j < adapter->req_rx_queues; j++) {
2514                 data[i] = adapter->rx_stats_buffers[j].packets;
2515                 i++;
2516                 data[i] = adapter->rx_stats_buffers[j].bytes;
2517                 i++;
2518                 data[i] = adapter->rx_stats_buffers[j].interrupts;
2519                 i++;
2520         }
2521 }
2522
2523 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2524 {
2525         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2526
2527         return adapter->priv_flags;
2528 }
2529
2530 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2531 {
2532         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2533         bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2534
2535         if (which_maxes)
2536                 adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2537         else
2538                 adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2539
2540         return 0;
2541 }
2542 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2543         .get_drvinfo            = ibmvnic_get_drvinfo,
2544         .get_msglevel           = ibmvnic_get_msglevel,
2545         .set_msglevel           = ibmvnic_set_msglevel,
2546         .get_link               = ibmvnic_get_link,
2547         .get_ringparam          = ibmvnic_get_ringparam,
2548         .set_ringparam          = ibmvnic_set_ringparam,
2549         .get_channels           = ibmvnic_get_channels,
2550         .set_channels           = ibmvnic_set_channels,
2551         .get_strings            = ibmvnic_get_strings,
2552         .get_sset_count         = ibmvnic_get_sset_count,
2553         .get_ethtool_stats      = ibmvnic_get_ethtool_stats,
2554         .get_link_ksettings     = ibmvnic_get_link_ksettings,
2555         .get_priv_flags         = ibmvnic_get_priv_flags,
2556         .set_priv_flags         = ibmvnic_set_priv_flags,
2557 };
2558
2559 /* Routines for managing CRQs/sCRQs  */
2560
2561 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2562                                    struct ibmvnic_sub_crq_queue *scrq)
2563 {
2564         int rc;
2565
2566         if (scrq->irq) {
2567                 free_irq(scrq->irq, scrq);
2568                 irq_dispose_mapping(scrq->irq);
2569                 scrq->irq = 0;
2570         }
2571
2572         memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2573         atomic_set(&scrq->used, 0);
2574         scrq->cur = 0;
2575
2576         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2577                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2578         return rc;
2579 }
2580
2581 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2582 {
2583         int i, rc;
2584
2585         for (i = 0; i < adapter->req_tx_queues; i++) {
2586                 netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2587                 rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2588                 if (rc)
2589                         return rc;
2590         }
2591
2592         for (i = 0; i < adapter->req_rx_queues; i++) {
2593                 netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2594                 rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2595                 if (rc)
2596                         return rc;
2597         }
2598
2599         return rc;
2600 }
2601
2602 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2603                                   struct ibmvnic_sub_crq_queue *scrq,
2604                                   bool do_h_free)
2605 {
2606         struct device *dev = &adapter->vdev->dev;
2607         long rc;
2608
2609         netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2610
2611         if (do_h_free) {
2612                 /* Close the sub-crqs */
2613                 do {
2614                         rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2615                                                 adapter->vdev->unit_address,
2616                                                 scrq->crq_num);
2617                 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2618
2619                 if (rc) {
2620                         netdev_err(adapter->netdev,
2621                                    "Failed to release sub-CRQ %16lx, rc = %ld\n",
2622                                    scrq->crq_num, rc);
2623                 }
2624         }
2625
2626         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2627                          DMA_BIDIRECTIONAL);
2628         free_pages((unsigned long)scrq->msgs, 2);
2629         kfree(scrq);
2630 }
2631
2632 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2633                                                         *adapter)
2634 {
2635         struct device *dev = &adapter->vdev->dev;
2636         struct ibmvnic_sub_crq_queue *scrq;
2637         int rc;
2638
2639         scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2640         if (!scrq)
2641                 return NULL;
2642
2643         scrq->msgs =
2644                 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2645         if (!scrq->msgs) {
2646                 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2647                 goto zero_page_failed;
2648         }
2649
2650         scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2651                                          DMA_BIDIRECTIONAL);
2652         if (dma_mapping_error(dev, scrq->msg_token)) {
2653                 dev_warn(dev, "Couldn't map crq queue messages page\n");
2654                 goto map_failed;
2655         }
2656
2657         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2658                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2659
2660         if (rc == H_RESOURCE)
2661                 rc = ibmvnic_reset_crq(adapter);
2662
2663         if (rc == H_CLOSED) {
2664                 dev_warn(dev, "Partner adapter not ready, waiting.\n");
2665         } else if (rc) {
2666                 dev_warn(dev, "Error %d registering sub-crq\n", rc);
2667                 goto reg_failed;
2668         }
2669
2670         scrq->adapter = adapter;
2671         scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2672         spin_lock_init(&scrq->lock);
2673
2674         netdev_dbg(adapter->netdev,
2675                    "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2676                    scrq->crq_num, scrq->hw_irq, scrq->irq);
2677
2678         return scrq;
2679
2680 reg_failed:
2681         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2682                          DMA_BIDIRECTIONAL);
2683 map_failed:
2684         free_pages((unsigned long)scrq->msgs, 2);
2685 zero_page_failed:
2686         kfree(scrq);
2687
2688         return NULL;
2689 }
2690
2691 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2692 {
2693         int i;
2694
2695         if (adapter->tx_scrq) {
2696                 for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2697                         if (!adapter->tx_scrq[i])
2698                                 continue;
2699
2700                         netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2701                                    i);
2702                         if (adapter->tx_scrq[i]->irq) {
2703                                 free_irq(adapter->tx_scrq[i]->irq,
2704                                          adapter->tx_scrq[i]);
2705                                 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2706                                 adapter->tx_scrq[i]->irq = 0;
2707                         }
2708
2709                         release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2710                                               do_h_free);
2711                 }
2712
2713                 kfree(adapter->tx_scrq);
2714                 adapter->tx_scrq = NULL;
2715                 adapter->num_active_tx_scrqs = 0;
2716         }
2717
2718         if (adapter->rx_scrq) {
2719                 for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2720                         if (!adapter->rx_scrq[i])
2721                                 continue;
2722
2723                         netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2724                                    i);
2725                         if (adapter->rx_scrq[i]->irq) {
2726                                 free_irq(adapter->rx_scrq[i]->irq,
2727                                          adapter->rx_scrq[i]);
2728                                 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2729                                 adapter->rx_scrq[i]->irq = 0;
2730                         }
2731
2732                         release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2733                                               do_h_free);
2734                 }
2735
2736                 kfree(adapter->rx_scrq);
2737                 adapter->rx_scrq = NULL;
2738                 adapter->num_active_rx_scrqs = 0;
2739         }
2740 }
2741
2742 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2743                             struct ibmvnic_sub_crq_queue *scrq)
2744 {
2745         struct device *dev = &adapter->vdev->dev;
2746         unsigned long rc;
2747
2748         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2749                                 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2750         if (rc)
2751                 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2752                         scrq->hw_irq, rc);
2753         return rc;
2754 }
2755
2756 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2757                            struct ibmvnic_sub_crq_queue *scrq)
2758 {
2759         struct device *dev = &adapter->vdev->dev;
2760         unsigned long rc;
2761
2762         if (scrq->hw_irq > 0x100000000ULL) {
2763                 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2764                 return 1;
2765         }
2766
2767         if (adapter->resetting &&
2768             adapter->reset_reason == VNIC_RESET_MOBILITY) {
2769                 u64 val = (0xff000000) | scrq->hw_irq;
2770
2771                 rc = plpar_hcall_norets(H_EOI, val);
2772                 if (rc)
2773                         dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
2774                                 val, rc);
2775         }
2776
2777         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2778                                 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2779         if (rc)
2780                 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2781                         scrq->hw_irq, rc);
2782         return rc;
2783 }
2784
2785 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2786                                struct ibmvnic_sub_crq_queue *scrq)
2787 {
2788         struct device *dev = &adapter->vdev->dev;
2789         struct ibmvnic_tx_pool *tx_pool;
2790         struct ibmvnic_tx_buff *txbuff;
2791         union sub_crq *next;
2792         int index;
2793         int i, j;
2794         u8 *first;
2795
2796 restart_loop:
2797         while (pending_scrq(adapter, scrq)) {
2798                 unsigned int pool = scrq->pool_index;
2799                 int num_entries = 0;
2800
2801                 next = ibmvnic_next_scrq(adapter, scrq);
2802                 for (i = 0; i < next->tx_comp.num_comps; i++) {
2803                         if (next->tx_comp.rcs[i]) {
2804                                 dev_err(dev, "tx error %x\n",
2805                                         next->tx_comp.rcs[i]);
2806                                 continue;
2807                         }
2808                         index = be32_to_cpu(next->tx_comp.correlators[i]);
2809                         if (index & IBMVNIC_TSO_POOL_MASK) {
2810                                 tx_pool = &adapter->tso_pool[pool];
2811                                 index &= ~IBMVNIC_TSO_POOL_MASK;
2812                         } else {
2813                                 tx_pool = &adapter->tx_pool[pool];
2814                         }
2815
2816                         txbuff = &tx_pool->tx_buff[index];
2817
2818                         for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2819                                 if (!txbuff->data_dma[j])
2820                                         continue;
2821
2822                                 txbuff->data_dma[j] = 0;
2823                         }
2824                         /* if sub_crq was sent indirectly */
2825                         first = &txbuff->indir_arr[0].generic.first;
2826                         if (*first == IBMVNIC_CRQ_CMD) {
2827                                 dma_unmap_single(dev, txbuff->indir_dma,
2828                                                  sizeof(txbuff->indir_arr),
2829                                                  DMA_TO_DEVICE);
2830                                 *first = 0;
2831                         }
2832
2833                         if (txbuff->last_frag) {
2834                                 dev_kfree_skb_any(txbuff->skb);
2835                                 txbuff->skb = NULL;
2836                         }
2837
2838                         num_entries += txbuff->num_entries;
2839
2840                         tx_pool->free_map[tx_pool->producer_index] = index;
2841                         tx_pool->producer_index =
2842                                 (tx_pool->producer_index + 1) %
2843                                         tx_pool->num_buffers;
2844                 }
2845                 /* remove tx_comp scrq*/
2846                 next->tx_comp.first = 0;
2847
2848                 if (atomic_sub_return(num_entries, &scrq->used) <=
2849                     (adapter->req_tx_entries_per_subcrq / 2) &&
2850                     __netif_subqueue_stopped(adapter->netdev,
2851                                              scrq->pool_index)) {
2852                         netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2853                         netdev_dbg(adapter->netdev, "Started queue %d\n",
2854                                    scrq->pool_index);
2855                 }
2856         }
2857
2858         enable_scrq_irq(adapter, scrq);
2859
2860         if (pending_scrq(adapter, scrq)) {
2861                 disable_scrq_irq(adapter, scrq);
2862                 goto restart_loop;
2863         }
2864
2865         return 0;
2866 }
2867
2868 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2869 {
2870         struct ibmvnic_sub_crq_queue *scrq = instance;
2871         struct ibmvnic_adapter *adapter = scrq->adapter;
2872
2873         disable_scrq_irq(adapter, scrq);
2874         ibmvnic_complete_tx(adapter, scrq);
2875
2876         return IRQ_HANDLED;
2877 }
2878
2879 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2880 {
2881         struct ibmvnic_sub_crq_queue *scrq = instance;
2882         struct ibmvnic_adapter *adapter = scrq->adapter;
2883
2884         /* When booting a kdump kernel we can hit pending interrupts
2885          * prior to completing driver initialization.
2886          */
2887         if (unlikely(adapter->state != VNIC_OPEN))
2888                 return IRQ_NONE;
2889
2890         adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2891
2892         if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2893                 disable_scrq_irq(adapter, scrq);
2894                 __napi_schedule(&adapter->napi[scrq->scrq_num]);
2895         }
2896
2897         return IRQ_HANDLED;
2898 }
2899
2900 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2901 {
2902         struct device *dev = &adapter->vdev->dev;
2903         struct ibmvnic_sub_crq_queue *scrq;
2904         int i = 0, j = 0;
2905         int rc = 0;
2906
2907         for (i = 0; i < adapter->req_tx_queues; i++) {
2908                 netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2909                            i);
2910                 scrq = adapter->tx_scrq[i];
2911                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2912
2913                 if (!scrq->irq) {
2914                         rc = -EINVAL;
2915                         dev_err(dev, "Error mapping irq\n");
2916                         goto req_tx_irq_failed;
2917                 }
2918
2919                 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2920                                  0, "ibmvnic_tx", scrq);
2921
2922                 if (rc) {
2923                         dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2924                                 scrq->irq, rc);
2925                         irq_dispose_mapping(scrq->irq);
2926                         goto req_tx_irq_failed;
2927                 }
2928         }
2929
2930         for (i = 0; i < adapter->req_rx_queues; i++) {
2931                 netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2932                            i);
2933                 scrq = adapter->rx_scrq[i];
2934                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2935                 if (!scrq->irq) {
2936                         rc = -EINVAL;
2937                         dev_err(dev, "Error mapping irq\n");
2938                         goto req_rx_irq_failed;
2939                 }
2940                 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2941                                  0, "ibmvnic_rx", scrq);
2942                 if (rc) {
2943                         dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2944                                 scrq->irq, rc);
2945                         irq_dispose_mapping(scrq->irq);
2946                         goto req_rx_irq_failed;
2947                 }
2948         }
2949         return rc;
2950
2951 req_rx_irq_failed:
2952         for (j = 0; j < i; j++) {
2953                 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2954                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2955         }
2956         i = adapter->req_tx_queues;
2957 req_tx_irq_failed:
2958         for (j = 0; j < i; j++) {
2959                 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2960                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2961         }
2962         release_sub_crqs(adapter, 1);
2963         return rc;
2964 }
2965
2966 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2967 {
2968         struct device *dev = &adapter->vdev->dev;
2969         struct ibmvnic_sub_crq_queue **allqueues;
2970         int registered_queues = 0;
2971         int total_queues;
2972         int more = 0;
2973         int i;
2974
2975         total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2976
2977         allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2978         if (!allqueues)
2979                 return -1;
2980
2981         for (i = 0; i < total_queues; i++) {
2982                 allqueues[i] = init_sub_crq_queue(adapter);
2983                 if (!allqueues[i]) {
2984                         dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2985                         break;
2986                 }
2987                 registered_queues++;
2988         }
2989
2990         /* Make sure we were able to register the minimum number of queues */
2991         if (registered_queues <
2992             adapter->min_tx_queues + adapter->min_rx_queues) {
2993                 dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2994                 goto tx_failed;
2995         }
2996
2997         /* Distribute the failed allocated queues*/
2998         for (i = 0; i < total_queues - registered_queues + more ; i++) {
2999                 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
3000                 switch (i % 3) {
3001                 case 0:
3002                         if (adapter->req_rx_queues > adapter->min_rx_queues)
3003                                 adapter->req_rx_queues--;
3004                         else
3005                                 more++;
3006                         break;
3007                 case 1:
3008                         if (adapter->req_tx_queues > adapter->min_tx_queues)
3009                                 adapter->req_tx_queues--;
3010                         else
3011                                 more++;
3012                         break;
3013                 }
3014         }
3015
3016         adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3017                                    sizeof(*adapter->tx_scrq), GFP_KERNEL);
3018         if (!adapter->tx_scrq)
3019                 goto tx_failed;
3020
3021         for (i = 0; i < adapter->req_tx_queues; i++) {
3022                 adapter->tx_scrq[i] = allqueues[i];
3023                 adapter->tx_scrq[i]->pool_index = i;
3024                 adapter->num_active_tx_scrqs++;
3025         }
3026
3027         adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3028                                    sizeof(*adapter->rx_scrq), GFP_KERNEL);
3029         if (!adapter->rx_scrq)
3030                 goto rx_failed;
3031
3032         for (i = 0; i < adapter->req_rx_queues; i++) {
3033                 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3034                 adapter->rx_scrq[i]->scrq_num = i;
3035                 adapter->num_active_rx_scrqs++;
3036         }
3037
3038         kfree(allqueues);
3039         return 0;
3040
3041 rx_failed:
3042         kfree(adapter->tx_scrq);
3043         adapter->tx_scrq = NULL;
3044 tx_failed:
3045         for (i = 0; i < registered_queues; i++)
3046                 release_sub_crq_queue(adapter, allqueues[i], 1);
3047         kfree(allqueues);
3048         return -1;
3049 }
3050
3051 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
3052 {
3053         struct device *dev = &adapter->vdev->dev;
3054         union ibmvnic_crq crq;
3055         int max_entries;
3056
3057         if (!retry) {
3058                 /* Sub-CRQ entries are 32 byte long */
3059                 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3060
3061                 if (adapter->min_tx_entries_per_subcrq > entries_page ||
3062                     adapter->min_rx_add_entries_per_subcrq > entries_page) {
3063                         dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3064                         return;
3065                 }
3066
3067                 if (adapter->desired.mtu)
3068                         adapter->req_mtu = adapter->desired.mtu;
3069                 else
3070                         adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3071
3072                 if (!adapter->desired.tx_entries)
3073                         adapter->desired.tx_entries =
3074                                         adapter->max_tx_entries_per_subcrq;
3075                 if (!adapter->desired.rx_entries)
3076                         adapter->desired.rx_entries =
3077                                         adapter->max_rx_add_entries_per_subcrq;
3078
3079                 max_entries = IBMVNIC_MAX_LTB_SIZE /
3080                               (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3081
3082                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3083                         adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3084                         adapter->desired.tx_entries = max_entries;
3085                 }
3086
3087                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3088                         adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3089                         adapter->desired.rx_entries = max_entries;
3090                 }
3091
3092                 if (adapter->desired.tx_entries)
3093                         adapter->req_tx_entries_per_subcrq =
3094                                         adapter->desired.tx_entries;
3095                 else
3096                         adapter->req_tx_entries_per_subcrq =
3097                                         adapter->max_tx_entries_per_subcrq;
3098
3099                 if (adapter->desired.rx_entries)
3100                         adapter->req_rx_add_entries_per_subcrq =
3101                                         adapter->desired.rx_entries;
3102                 else
3103                         adapter->req_rx_add_entries_per_subcrq =
3104                                         adapter->max_rx_add_entries_per_subcrq;
3105
3106                 if (adapter->desired.tx_queues)
3107                         adapter->req_tx_queues =
3108                                         adapter->desired.tx_queues;
3109                 else
3110                         adapter->req_tx_queues =
3111                                         adapter->opt_tx_comp_sub_queues;
3112
3113                 if (adapter->desired.rx_queues)
3114                         adapter->req_rx_queues =
3115                                         adapter->desired.rx_queues;
3116                 else
3117                         adapter->req_rx_queues =
3118                                         adapter->opt_rx_comp_queues;
3119
3120                 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3121         }
3122
3123         memset(&crq, 0, sizeof(crq));
3124         crq.request_capability.first = IBMVNIC_CRQ_CMD;
3125         crq.request_capability.cmd = REQUEST_CAPABILITY;
3126
3127         crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3128         crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3129         atomic_inc(&adapter->running_cap_crqs);
3130         ibmvnic_send_crq(adapter, &crq);
3131
3132         crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3133         crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3134         atomic_inc(&adapter->running_cap_crqs);
3135         ibmvnic_send_crq(adapter, &crq);
3136
3137         crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3138         crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3139         atomic_inc(&adapter->running_cap_crqs);
3140         ibmvnic_send_crq(adapter, &crq);
3141
3142         crq.request_capability.capability =
3143             cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3144         crq.request_capability.number =
3145             cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3146         atomic_inc(&adapter->running_cap_crqs);
3147         ibmvnic_send_crq(adapter, &crq);
3148
3149         crq.request_capability.capability =
3150             cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3151         crq.request_capability.number =
3152             cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3153         atomic_inc(&adapter->running_cap_crqs);
3154         ibmvnic_send_crq(adapter, &crq);
3155
3156         crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3157         crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3158         atomic_inc(&adapter->running_cap_crqs);
3159         ibmvnic_send_crq(adapter, &crq);
3160
3161         if (adapter->netdev->flags & IFF_PROMISC) {
3162                 if (adapter->promisc_supported) {
3163                         crq.request_capability.capability =
3164                             cpu_to_be16(PROMISC_REQUESTED);
3165                         crq.request_capability.number = cpu_to_be64(1);
3166                         atomic_inc(&adapter->running_cap_crqs);
3167                         ibmvnic_send_crq(adapter, &crq);
3168                 }
3169         } else {
3170                 crq.request_capability.capability =
3171                     cpu_to_be16(PROMISC_REQUESTED);
3172                 crq.request_capability.number = cpu_to_be64(0);
3173                 atomic_inc(&adapter->running_cap_crqs);
3174                 ibmvnic_send_crq(adapter, &crq);
3175         }
3176 }
3177
3178 static int pending_scrq(struct ibmvnic_adapter *adapter,
3179                         struct ibmvnic_sub_crq_queue *scrq)
3180 {
3181         union sub_crq *entry = &scrq->msgs[scrq->cur];
3182
3183         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3184                 return 1;
3185         else
3186                 return 0;
3187 }
3188
3189 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3190                                         struct ibmvnic_sub_crq_queue *scrq)
3191 {
3192         union sub_crq *entry;
3193         unsigned long flags;
3194
3195         spin_lock_irqsave(&scrq->lock, flags);
3196         entry = &scrq->msgs[scrq->cur];
3197         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3198                 if (++scrq->cur == scrq->size)
3199                         scrq->cur = 0;
3200         } else {
3201                 entry = NULL;
3202         }
3203         spin_unlock_irqrestore(&scrq->lock, flags);
3204
3205         return entry;
3206 }
3207
3208 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3209 {
3210         struct ibmvnic_crq_queue *queue = &adapter->crq;
3211         union ibmvnic_crq *crq;
3212
3213         crq = &queue->msgs[queue->cur];
3214         if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3215                 if (++queue->cur == queue->size)
3216                         queue->cur = 0;
3217         } else {
3218                 crq = NULL;
3219         }
3220
3221         return crq;
3222 }
3223
3224 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3225 {
3226         switch (rc) {
3227         case H_PARAMETER:
3228                 dev_warn_ratelimited(dev,
3229                                      "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3230                                      func, rc);
3231                 break;
3232         case H_CLOSED:
3233                 dev_warn_ratelimited(dev,
3234                                      "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3235                                      func, rc);
3236                 break;
3237         default:
3238                 dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3239                 break;
3240         }
3241 }
3242
3243 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3244                        union sub_crq *sub_crq)
3245 {
3246         unsigned int ua = adapter->vdev->unit_address;
3247         struct device *dev = &adapter->vdev->dev;
3248         u64 *u64_crq = (u64 *)sub_crq;
3249         int rc;
3250
3251         netdev_dbg(adapter->netdev,
3252                    "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3253                    (unsigned long int)cpu_to_be64(remote_handle),
3254                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3255                    (unsigned long int)cpu_to_be64(u64_crq[1]),
3256                    (unsigned long int)cpu_to_be64(u64_crq[2]),
3257                    (unsigned long int)cpu_to_be64(u64_crq[3]));
3258
3259         /* Make sure the hypervisor sees the complete request */
3260         mb();
3261
3262         rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3263                                 cpu_to_be64(remote_handle),
3264                                 cpu_to_be64(u64_crq[0]),
3265                                 cpu_to_be64(u64_crq[1]),
3266                                 cpu_to_be64(u64_crq[2]),
3267                                 cpu_to_be64(u64_crq[3]));
3268
3269         if (rc)
3270                 print_subcrq_error(dev, rc, __func__);
3271
3272         return rc;
3273 }
3274
3275 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3276                                 u64 remote_handle, u64 ioba, u64 num_entries)
3277 {
3278         unsigned int ua = adapter->vdev->unit_address;
3279         struct device *dev = &adapter->vdev->dev;
3280         int rc;
3281
3282         /* Make sure the hypervisor sees the complete request */
3283         mb();
3284         rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3285                                 cpu_to_be64(remote_handle),
3286                                 ioba, num_entries);
3287
3288         if (rc)
3289                 print_subcrq_error(dev, rc, __func__);
3290
3291         return rc;
3292 }
3293
3294 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3295                             union ibmvnic_crq *crq)
3296 {
3297         unsigned int ua = adapter->vdev->unit_address;
3298         struct device *dev = &adapter->vdev->dev;
3299         u64 *u64_crq = (u64 *)crq;
3300         int rc;
3301
3302         netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3303                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3304                    (unsigned long int)cpu_to_be64(u64_crq[1]));
3305
3306         if (!adapter->crq.active &&
3307             crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3308                 dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3309                 return -EINVAL;
3310         }
3311
3312         /* Make sure the hypervisor sees the complete request */
3313         mb();
3314
3315         rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3316                                 cpu_to_be64(u64_crq[0]),
3317                                 cpu_to_be64(u64_crq[1]));
3318
3319         if (rc) {
3320                 if (rc == H_CLOSED) {
3321                         dev_warn(dev, "CRQ Queue closed\n");
3322                         if (adapter->resetting)
3323                                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3324                 }
3325
3326                 dev_warn(dev, "Send error (rc=%d)\n", rc);
3327         }
3328
3329         return rc;
3330 }
3331
3332 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3333 {
3334         union ibmvnic_crq crq;
3335
3336         memset(&crq, 0, sizeof(crq));
3337         crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3338         crq.generic.cmd = IBMVNIC_CRQ_INIT;
3339         netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3340
3341         return ibmvnic_send_crq(adapter, &crq);
3342 }
3343
3344 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3345 {
3346         union ibmvnic_crq crq;
3347
3348         memset(&crq, 0, sizeof(crq));
3349         crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3350         crq.version_exchange.cmd = VERSION_EXCHANGE;
3351         crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3352
3353         return ibmvnic_send_crq(adapter, &crq);
3354 }
3355
3356 struct vnic_login_client_data {
3357         u8      type;
3358         __be16  len;
3359         char    name[];
3360 } __packed;
3361
3362 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3363 {
3364         int len;
3365
3366         /* Calculate the amount of buffer space needed for the
3367          * vnic client data in the login buffer. There are four entries,
3368          * OS name, LPAR name, device name, and a null last entry.
3369          */
3370         len = 4 * sizeof(struct vnic_login_client_data);
3371         len += 6; /* "Linux" plus NULL */
3372         len += strlen(utsname()->nodename) + 1;
3373         len += strlen(adapter->netdev->name) + 1;
3374
3375         return len;
3376 }
3377
3378 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3379                                  struct vnic_login_client_data *vlcd)
3380 {
3381         const char *os_name = "Linux";
3382         int len;
3383
3384         /* Type 1 - LPAR OS */
3385         vlcd->type = 1;
3386         len = strlen(os_name) + 1;
3387         vlcd->len = cpu_to_be16(len);
3388         strncpy(vlcd->name, os_name, len);
3389         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3390
3391         /* Type 2 - LPAR name */
3392         vlcd->type = 2;
3393         len = strlen(utsname()->nodename) + 1;
3394         vlcd->len = cpu_to_be16(len);
3395         strncpy(vlcd->name, utsname()->nodename, len);
3396         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3397
3398         /* Type 3 - device name */
3399         vlcd->type = 3;
3400         len = strlen(adapter->netdev->name) + 1;
3401         vlcd->len = cpu_to_be16(len);
3402         strncpy(vlcd->name, adapter->netdev->name, len);
3403 }
3404
3405 static int send_login(struct ibmvnic_adapter *adapter)
3406 {
3407         struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3408         struct ibmvnic_login_buffer *login_buffer;
3409         struct device *dev = &adapter->vdev->dev;
3410         dma_addr_t rsp_buffer_token;
3411         dma_addr_t buffer_token;
3412         size_t rsp_buffer_size;
3413         union ibmvnic_crq crq;
3414         size_t buffer_size;
3415         __be64 *tx_list_p;
3416         __be64 *rx_list_p;
3417         int client_data_len;
3418         struct vnic_login_client_data *vlcd;
3419         int i;
3420
3421         if (!adapter->tx_scrq || !adapter->rx_scrq) {
3422                 netdev_err(adapter->netdev,
3423                            "RX or TX queues are not allocated, device login failed\n");
3424                 return -1;
3425         }
3426
3427         release_login_rsp_buffer(adapter);
3428         client_data_len = vnic_client_data_len(adapter);
3429
3430         buffer_size =
3431             sizeof(struct ibmvnic_login_buffer) +
3432             sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3433             client_data_len;
3434
3435         login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3436         if (!login_buffer)
3437                 goto buf_alloc_failed;
3438
3439         buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3440                                       DMA_TO_DEVICE);
3441         if (dma_mapping_error(dev, buffer_token)) {
3442                 dev_err(dev, "Couldn't map login buffer\n");
3443                 goto buf_map_failed;
3444         }
3445
3446         rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3447                           sizeof(u64) * adapter->req_tx_queues +
3448                           sizeof(u64) * adapter->req_rx_queues +
3449                           sizeof(u64) * adapter->req_rx_queues +
3450                           sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3451
3452         login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3453         if (!login_rsp_buffer)
3454                 goto buf_rsp_alloc_failed;
3455
3456         rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3457                                           rsp_buffer_size, DMA_FROM_DEVICE);
3458         if (dma_mapping_error(dev, rsp_buffer_token)) {
3459                 dev_err(dev, "Couldn't map login rsp buffer\n");
3460                 goto buf_rsp_map_failed;
3461         }
3462
3463         adapter->login_buf = login_buffer;
3464         adapter->login_buf_token = buffer_token;
3465         adapter->login_buf_sz = buffer_size;
3466         adapter->login_rsp_buf = login_rsp_buffer;
3467         adapter->login_rsp_buf_token = rsp_buffer_token;
3468         adapter->login_rsp_buf_sz = rsp_buffer_size;
3469
3470         login_buffer->len = cpu_to_be32(buffer_size);
3471         login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3472         login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3473         login_buffer->off_txcomp_subcrqs =
3474             cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3475         login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3476         login_buffer->off_rxcomp_subcrqs =
3477             cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3478                         sizeof(u64) * adapter->req_tx_queues);
3479         login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3480         login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3481
3482         tx_list_p = (__be64 *)((char *)login_buffer +
3483                                       sizeof(struct ibmvnic_login_buffer));
3484         rx_list_p = (__be64 *)((char *)login_buffer +
3485                                       sizeof(struct ibmvnic_login_buffer) +
3486                                       sizeof(u64) * adapter->req_tx_queues);
3487
3488         for (i = 0; i < adapter->req_tx_queues; i++) {
3489                 if (adapter->tx_scrq[i]) {
3490                         tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3491                                                    crq_num);
3492                 }
3493         }
3494
3495         for (i = 0; i < adapter->req_rx_queues; i++) {
3496                 if (adapter->rx_scrq[i]) {
3497                         rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3498                                                    crq_num);
3499                 }
3500         }
3501
3502         /* Insert vNIC login client data */
3503         vlcd = (struct vnic_login_client_data *)
3504                 ((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3505         login_buffer->client_data_offset =
3506                         cpu_to_be32((char *)vlcd - (char *)login_buffer);
3507         login_buffer->client_data_len = cpu_to_be32(client_data_len);
3508
3509         vnic_add_client_data(adapter, vlcd);
3510
3511         netdev_dbg(adapter->netdev, "Login Buffer:\n");
3512         for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3513                 netdev_dbg(adapter->netdev, "%016lx\n",
3514                            ((unsigned long int *)(adapter->login_buf))[i]);
3515         }
3516
3517         memset(&crq, 0, sizeof(crq));
3518         crq.login.first = IBMVNIC_CRQ_CMD;
3519         crq.login.cmd = LOGIN;
3520         crq.login.ioba = cpu_to_be32(buffer_token);
3521         crq.login.len = cpu_to_be32(buffer_size);
3522         ibmvnic_send_crq(adapter, &crq);
3523
3524         return 0;
3525
3526 buf_rsp_map_failed:
3527         kfree(login_rsp_buffer);
3528 buf_rsp_alloc_failed:
3529         dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3530 buf_map_failed:
3531         kfree(login_buffer);
3532 buf_alloc_failed:
3533         return -1;
3534 }
3535
3536 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3537                             u32 len, u8 map_id)
3538 {
3539         union ibmvnic_crq crq;
3540
3541         memset(&crq, 0, sizeof(crq));
3542         crq.request_map.first = IBMVNIC_CRQ_CMD;
3543         crq.request_map.cmd = REQUEST_MAP;
3544         crq.request_map.map_id = map_id;
3545         crq.request_map.ioba = cpu_to_be32(addr);
3546         crq.request_map.len = cpu_to_be32(len);
3547         return ibmvnic_send_crq(adapter, &crq);
3548 }
3549
3550 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3551 {
3552         union ibmvnic_crq crq;
3553
3554         memset(&crq, 0, sizeof(crq));
3555         crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3556         crq.request_unmap.cmd = REQUEST_UNMAP;
3557         crq.request_unmap.map_id = map_id;
3558         return ibmvnic_send_crq(adapter, &crq);
3559 }
3560
3561 static void send_map_query(struct ibmvnic_adapter *adapter)
3562 {
3563         union ibmvnic_crq crq;
3564
3565         memset(&crq, 0, sizeof(crq));
3566         crq.query_map.first = IBMVNIC_CRQ_CMD;
3567         crq.query_map.cmd = QUERY_MAP;
3568         ibmvnic_send_crq(adapter, &crq);
3569 }
3570
3571 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3572 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3573 {
3574         union ibmvnic_crq crq;
3575
3576         atomic_set(&adapter->running_cap_crqs, 0);
3577         memset(&crq, 0, sizeof(crq));
3578         crq.query_capability.first = IBMVNIC_CRQ_CMD;
3579         crq.query_capability.cmd = QUERY_CAPABILITY;
3580
3581         crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3582         atomic_inc(&adapter->running_cap_crqs);
3583         ibmvnic_send_crq(adapter, &crq);
3584
3585         crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3586         atomic_inc(&adapter->running_cap_crqs);
3587         ibmvnic_send_crq(adapter, &crq);
3588
3589         crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3590         atomic_inc(&adapter->running_cap_crqs);
3591         ibmvnic_send_crq(adapter, &crq);
3592
3593         crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3594         atomic_inc(&adapter->running_cap_crqs);
3595         ibmvnic_send_crq(adapter, &crq);
3596
3597         crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3598         atomic_inc(&adapter->running_cap_crqs);
3599         ibmvnic_send_crq(adapter, &crq);
3600
3601         crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3602         atomic_inc(&adapter->running_cap_crqs);
3603         ibmvnic_send_crq(adapter, &crq);
3604
3605         crq.query_capability.capability =
3606             cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3607         atomic_inc(&adapter->running_cap_crqs);
3608         ibmvnic_send_crq(adapter, &crq);
3609
3610         crq.query_capability.capability =
3611             cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3612         atomic_inc(&adapter->running_cap_crqs);
3613         ibmvnic_send_crq(adapter, &crq);
3614
3615         crq.query_capability.capability =
3616             cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3617         atomic_inc(&adapter->running_cap_crqs);
3618         ibmvnic_send_crq(adapter, &crq);
3619
3620         crq.query_capability.capability =
3621             cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3622         atomic_inc(&adapter->running_cap_crqs);
3623         ibmvnic_send_crq(adapter, &crq);
3624
3625         crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3626         atomic_inc(&adapter->running_cap_crqs);
3627         ibmvnic_send_crq(adapter, &crq);
3628
3629         crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3630         atomic_inc(&adapter->running_cap_crqs);
3631         ibmvnic_send_crq(adapter, &crq);
3632
3633         crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3634         atomic_inc(&adapter->running_cap_crqs);
3635         ibmvnic_send_crq(adapter, &crq);
3636
3637         crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3638         atomic_inc(&adapter->running_cap_crqs);
3639         ibmvnic_send_crq(adapter, &crq);
3640
3641         crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3642         atomic_inc(&adapter->running_cap_crqs);
3643         ibmvnic_send_crq(adapter, &crq);
3644
3645         crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3646         atomic_inc(&adapter->running_cap_crqs);
3647         ibmvnic_send_crq(adapter, &crq);
3648
3649         crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3650         atomic_inc(&adapter->running_cap_crqs);
3651         ibmvnic_send_crq(adapter, &crq);
3652
3653         crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3654         atomic_inc(&adapter->running_cap_crqs);
3655         ibmvnic_send_crq(adapter, &crq);
3656
3657         crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3658         atomic_inc(&adapter->running_cap_crqs);
3659         ibmvnic_send_crq(adapter, &crq);
3660
3661         crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3662         atomic_inc(&adapter->running_cap_crqs);
3663         ibmvnic_send_crq(adapter, &crq);
3664
3665         crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3666         atomic_inc(&adapter->running_cap_crqs);
3667         ibmvnic_send_crq(adapter, &crq);
3668
3669         crq.query_capability.capability =
3670                         cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3671         atomic_inc(&adapter->running_cap_crqs);
3672         ibmvnic_send_crq(adapter, &crq);
3673
3674         crq.query_capability.capability =
3675                         cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3676         atomic_inc(&adapter->running_cap_crqs);
3677         ibmvnic_send_crq(adapter, &crq);
3678
3679         crq.query_capability.capability =
3680                         cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3681         atomic_inc(&adapter->running_cap_crqs);
3682         ibmvnic_send_crq(adapter, &crq);
3683
3684         crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3685         atomic_inc(&adapter->running_cap_crqs);
3686         ibmvnic_send_crq(adapter, &crq);
3687 }
3688
3689 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3690                                 struct ibmvnic_adapter *adapter)
3691 {
3692         struct device *dev = &adapter->vdev->dev;
3693
3694         if (crq->get_vpd_size_rsp.rc.code) {
3695                 dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3696                         crq->get_vpd_size_rsp.rc.code);
3697                 complete(&adapter->fw_done);
3698                 return;
3699         }
3700
3701         adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3702         complete(&adapter->fw_done);
3703 }
3704
3705 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3706                            struct ibmvnic_adapter *adapter)
3707 {
3708         struct device *dev = &adapter->vdev->dev;
3709         unsigned char *substr = NULL;
3710         u8 fw_level_len = 0;
3711
3712         memset(adapter->fw_version, 0, 32);
3713
3714         dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3715                          DMA_FROM_DEVICE);
3716
3717         if (crq->get_vpd_rsp.rc.code) {
3718                 dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3719                         crq->get_vpd_rsp.rc.code);
3720                 goto complete;
3721         }
3722
3723         /* get the position of the firmware version info
3724          * located after the ASCII 'RM' substring in the buffer
3725          */
3726         substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3727         if (!substr) {
3728                 dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3729                 goto complete;
3730         }
3731
3732         /* get length of firmware level ASCII substring */
3733         if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3734                 fw_level_len = *(substr + 2);
3735         } else {
3736                 dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3737                 goto complete;
3738         }
3739
3740         /* copy firmware version string from vpd into adapter */
3741         if ((substr + 3 + fw_level_len) <
3742             (adapter->vpd->buff + adapter->vpd->len)) {
3743                 strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3744         } else {
3745                 dev_info(dev, "FW substr extrapolated VPD buff\n");
3746         }
3747
3748 complete:
3749         if (adapter->fw_version[0] == '\0')
3750                 strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3751         complete(&adapter->fw_done);
3752 }
3753
3754 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3755 {
3756         struct device *dev = &adapter->vdev->dev;
3757         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3758         union ibmvnic_crq crq;
3759         int i;
3760
3761         dma_unmap_single(dev, adapter->ip_offload_tok,
3762                          sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3763
3764         netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3765         for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3766                 netdev_dbg(adapter->netdev, "%016lx\n",
3767                            ((unsigned long int *)(buf))[i]);
3768
3769         netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3770         netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3771         netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3772                    buf->tcp_ipv4_chksum);
3773         netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3774                    buf->tcp_ipv6_chksum);
3775         netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3776                    buf->udp_ipv4_chksum);
3777         netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3778                    buf->udp_ipv6_chksum);
3779         netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3780                    buf->large_tx_ipv4);
3781         netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3782                    buf->large_tx_ipv6);
3783         netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3784                    buf->large_rx_ipv4);
3785         netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3786                    buf->large_rx_ipv6);
3787         netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3788                    buf->max_ipv4_header_size);
3789         netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3790                    buf->max_ipv6_header_size);
3791         netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3792                    buf->max_tcp_header_size);
3793         netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3794                    buf->max_udp_header_size);
3795         netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3796                    buf->max_large_tx_size);
3797         netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3798                    buf->max_large_rx_size);
3799         netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3800                    buf->ipv6_extension_header);
3801         netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3802                    buf->tcp_pseudosum_req);
3803         netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3804                    buf->num_ipv6_ext_headers);
3805         netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3806                    buf->off_ipv6_ext_headers);
3807
3808         adapter->ip_offload_ctrl_tok =
3809             dma_map_single(dev, &adapter->ip_offload_ctrl,
3810                            sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3811
3812         if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3813                 dev_err(dev, "Couldn't map ip offload control buffer\n");
3814                 return;
3815         }
3816
3817         adapter->ip_offload_ctrl.len =
3818             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3819         adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3820         adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3821         adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3822         adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3823         adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3824         adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3825         adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3826         adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3827         adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3828
3829         /* large_rx disabled for now, additional features needed */
3830         adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3831         adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3832
3833         adapter->netdev->features = NETIF_F_SG | NETIF_F_GSO;
3834
3835         if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3836                 adapter->netdev->features |= NETIF_F_IP_CSUM;
3837
3838         if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3839                 adapter->netdev->features |= NETIF_F_IPV6_CSUM;
3840
3841         if ((adapter->netdev->features &
3842             (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3843                 adapter->netdev->features |= NETIF_F_RXCSUM;
3844
3845         if (buf->large_tx_ipv4)
3846                 adapter->netdev->features |= NETIF_F_TSO;
3847         if (buf->large_tx_ipv6)
3848                 adapter->netdev->features |= NETIF_F_TSO6;
3849
3850         adapter->netdev->hw_features |= adapter->netdev->features;
3851
3852         memset(&crq, 0, sizeof(crq));
3853         crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3854         crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3855         crq.control_ip_offload.len =
3856             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3857         crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3858         ibmvnic_send_crq(adapter, &crq);
3859 }
3860
3861 static const char *ibmvnic_fw_err_cause(u16 cause)
3862 {
3863         switch (cause) {
3864         case ADAPTER_PROBLEM:
3865                 return "adapter problem";
3866         case BUS_PROBLEM:
3867                 return "bus problem";
3868         case FW_PROBLEM:
3869                 return "firmware problem";
3870         case DD_PROBLEM:
3871                 return "device driver problem";
3872         case EEH_RECOVERY:
3873                 return "EEH recovery";
3874         case FW_UPDATED:
3875                 return "firmware updated";
3876         case LOW_MEMORY:
3877                 return "low Memory";
3878         default:
3879                 return "unknown";
3880         }
3881 }
3882
3883 static void handle_error_indication(union ibmvnic_crq *crq,
3884                                     struct ibmvnic_adapter *adapter)
3885 {
3886         struct device *dev = &adapter->vdev->dev;
3887         u16 cause;
3888
3889         cause = be16_to_cpu(crq->error_indication.error_cause);
3890
3891         dev_warn_ratelimited(dev,
3892                              "Firmware reports %serror, cause: %s. Starting recovery...\n",
3893                              crq->error_indication.flags
3894                                 & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3895                              ibmvnic_fw_err_cause(cause));
3896
3897         if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3898                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3899         else
3900                 ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3901 }
3902
3903 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
3904                                  struct ibmvnic_adapter *adapter)
3905 {
3906         struct net_device *netdev = adapter->netdev;
3907         struct device *dev = &adapter->vdev->dev;
3908         long rc;
3909
3910         rc = crq->change_mac_addr_rsp.rc.code;
3911         if (rc) {
3912                 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3913                 goto out;
3914         }
3915         memcpy(netdev->dev_addr, &crq->change_mac_addr_rsp.mac_addr[0],
3916                ETH_ALEN);
3917 out:
3918         complete(&adapter->fw_done);
3919         return rc;
3920 }
3921
3922 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3923                                    struct ibmvnic_adapter *adapter)
3924 {
3925         struct device *dev = &adapter->vdev->dev;
3926         u64 *req_value;
3927         char *name;
3928
3929         atomic_dec(&adapter->running_cap_crqs);
3930         switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3931         case REQ_TX_QUEUES:
3932                 req_value = &adapter->req_tx_queues;
3933                 name = "tx";
3934                 break;
3935         case REQ_RX_QUEUES:
3936                 req_value = &adapter->req_rx_queues;
3937                 name = "rx";
3938                 break;
3939         case REQ_RX_ADD_QUEUES:
3940                 req_value = &adapter->req_rx_add_queues;
3941                 name = "rx_add";
3942                 break;
3943         case REQ_TX_ENTRIES_PER_SUBCRQ:
3944                 req_value = &adapter->req_tx_entries_per_subcrq;
3945                 name = "tx_entries_per_subcrq";
3946                 break;
3947         case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3948                 req_value = &adapter->req_rx_add_entries_per_subcrq;
3949                 name = "rx_add_entries_per_subcrq";
3950                 break;
3951         case REQ_MTU:
3952                 req_value = &adapter->req_mtu;
3953                 name = "mtu";
3954                 break;
3955         case PROMISC_REQUESTED:
3956                 req_value = &adapter->promisc;
3957                 name = "promisc";
3958                 break;
3959         default:
3960                 dev_err(dev, "Got invalid cap request rsp %d\n",
3961                         crq->request_capability.capability);
3962                 return;
3963         }
3964
3965         switch (crq->request_capability_rsp.rc.code) {
3966         case SUCCESS:
3967                 break;
3968         case PARTIALSUCCESS:
3969                 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3970                          *req_value,
3971                          (long int)be64_to_cpu(crq->request_capability_rsp.
3972                                                number), name);
3973
3974                 if (be16_to_cpu(crq->request_capability_rsp.capability) ==
3975                     REQ_MTU) {
3976                         pr_err("mtu of %llu is not supported. Reverting.\n",
3977                                *req_value);
3978                         *req_value = adapter->fallback.mtu;
3979                 } else {
3980                         *req_value =
3981                                 be64_to_cpu(crq->request_capability_rsp.number);
3982                 }
3983
3984                 ibmvnic_send_req_caps(adapter, 1);
3985                 return;
3986         default:
3987                 dev_err(dev, "Error %d in request cap rsp\n",
3988                         crq->request_capability_rsp.rc.code);
3989                 return;
3990         }
3991
3992         /* Done receiving requested capabilities, query IP offload support */
3993         if (atomic_read(&adapter->running_cap_crqs) == 0) {
3994                 union ibmvnic_crq newcrq;
3995                 int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
3996                 struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
3997                     &adapter->ip_offload_buf;
3998
3999                 adapter->wait_capability = false;
4000                 adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
4001                                                          buf_sz,
4002                                                          DMA_FROM_DEVICE);
4003
4004                 if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
4005                         if (!firmware_has_feature(FW_FEATURE_CMO))
4006                                 dev_err(dev, "Couldn't map offload buffer\n");
4007                         return;
4008                 }
4009
4010                 memset(&newcrq, 0, sizeof(newcrq));
4011                 newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
4012                 newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
4013                 newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
4014                 newcrq.query_ip_offload.ioba =
4015                     cpu_to_be32(adapter->ip_offload_tok);
4016
4017                 ibmvnic_send_crq(adapter, &newcrq);
4018         }
4019 }
4020
4021 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4022                             struct ibmvnic_adapter *adapter)
4023 {
4024         struct device *dev = &adapter->vdev->dev;
4025         struct net_device *netdev = adapter->netdev;
4026         struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4027         struct ibmvnic_login_buffer *login = adapter->login_buf;
4028         int i;
4029
4030         dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4031                          DMA_TO_DEVICE);
4032         dma_unmap_single(dev, adapter->login_rsp_buf_token,
4033                          adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4034
4035         /* If the number of queues requested can't be allocated by the
4036          * server, the login response will return with code 1. We will need
4037          * to resend the login buffer with fewer queues requested.
4038          */
4039         if (login_rsp_crq->generic.rc.code) {
4040                 adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4041                 complete(&adapter->init_done);
4042                 return 0;
4043         }
4044
4045         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4046
4047         netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4048         for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4049                 netdev_dbg(adapter->netdev, "%016lx\n",
4050                            ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4051         }
4052
4053         /* Sanity checks */
4054         if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4055             (be32_to_cpu(login->num_rxcomp_subcrqs) *
4056              adapter->req_rx_add_queues !=
4057              be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4058                 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4059                 ibmvnic_remove(adapter->vdev);
4060                 return -EIO;
4061         }
4062         release_login_buffer(adapter);
4063         complete(&adapter->init_done);
4064
4065         return 0;
4066 }
4067
4068 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4069                                      struct ibmvnic_adapter *adapter)
4070 {
4071         struct device *dev = &adapter->vdev->dev;
4072         long rc;
4073
4074         rc = crq->request_unmap_rsp.rc.code;
4075         if (rc)
4076                 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4077 }
4078
4079 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4080                                  struct ibmvnic_adapter *adapter)
4081 {
4082         struct net_device *netdev = adapter->netdev;
4083         struct device *dev = &adapter->vdev->dev;
4084         long rc;
4085
4086         rc = crq->query_map_rsp.rc.code;
4087         if (rc) {
4088                 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4089                 return;
4090         }
4091         netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4092                    crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4093                    crq->query_map_rsp.free_pages);
4094 }
4095
4096 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4097                                  struct ibmvnic_adapter *adapter)
4098 {
4099         struct net_device *netdev = adapter->netdev;
4100         struct device *dev = &adapter->vdev->dev;
4101         long rc;
4102
4103         atomic_dec(&adapter->running_cap_crqs);
4104         netdev_dbg(netdev, "Outstanding queries: %d\n",
4105                    atomic_read(&adapter->running_cap_crqs));
4106         rc = crq->query_capability.rc.code;
4107         if (rc) {
4108                 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4109                 goto out;
4110         }
4111
4112         switch (be16_to_cpu(crq->query_capability.capability)) {
4113         case MIN_TX_QUEUES:
4114                 adapter->min_tx_queues =
4115                     be64_to_cpu(crq->query_capability.number);
4116                 netdev_dbg(netdev, "min_tx_queues = %lld\n",
4117                            adapter->min_tx_queues);
4118                 break;
4119         case MIN_RX_QUEUES:
4120                 adapter->min_rx_queues =
4121                     be64_to_cpu(crq->query_capability.number);
4122                 netdev_dbg(netdev, "min_rx_queues = %lld\n",
4123                            adapter->min_rx_queues);
4124                 break;
4125         case MIN_RX_ADD_QUEUES:
4126                 adapter->min_rx_add_queues =
4127                     be64_to_cpu(crq->query_capability.number);
4128                 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4129                            adapter->min_rx_add_queues);
4130                 break;
4131         case MAX_TX_QUEUES:
4132                 adapter->max_tx_queues =
4133                     be64_to_cpu(crq->query_capability.number);
4134                 netdev_dbg(netdev, "max_tx_queues = %lld\n",
4135                            adapter->max_tx_queues);
4136                 break;
4137         case MAX_RX_QUEUES:
4138                 adapter->max_rx_queues =
4139                     be64_to_cpu(crq->query_capability.number);
4140                 netdev_dbg(netdev, "max_rx_queues = %lld\n",
4141                            adapter->max_rx_queues);
4142                 break;
4143         case MAX_RX_ADD_QUEUES:
4144                 adapter->max_rx_add_queues =
4145                     be64_to_cpu(crq->query_capability.number);
4146                 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4147                            adapter->max_rx_add_queues);
4148                 break;
4149         case MIN_TX_ENTRIES_PER_SUBCRQ:
4150                 adapter->min_tx_entries_per_subcrq =
4151                     be64_to_cpu(crq->query_capability.number);
4152                 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4153                            adapter->min_tx_entries_per_subcrq);
4154                 break;
4155         case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4156                 adapter->min_rx_add_entries_per_subcrq =
4157                     be64_to_cpu(crq->query_capability.number);
4158                 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4159                            adapter->min_rx_add_entries_per_subcrq);
4160                 break;
4161         case MAX_TX_ENTRIES_PER_SUBCRQ:
4162                 adapter->max_tx_entries_per_subcrq =
4163                     be64_to_cpu(crq->query_capability.number);
4164                 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4165                            adapter->max_tx_entries_per_subcrq);
4166                 break;
4167         case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4168                 adapter->max_rx_add_entries_per_subcrq =
4169                     be64_to_cpu(crq->query_capability.number);
4170                 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4171                            adapter->max_rx_add_entries_per_subcrq);
4172                 break;
4173         case TCP_IP_OFFLOAD:
4174                 adapter->tcp_ip_offload =
4175                     be64_to_cpu(crq->query_capability.number);
4176                 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4177                            adapter->tcp_ip_offload);
4178                 break;
4179         case PROMISC_SUPPORTED:
4180                 adapter->promisc_supported =
4181                     be64_to_cpu(crq->query_capability.number);
4182                 netdev_dbg(netdev, "promisc_supported = %lld\n",
4183                            adapter->promisc_supported);
4184                 break;
4185         case MIN_MTU:
4186                 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4187                 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4188                 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4189                 break;
4190         case MAX_MTU:
4191                 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4192                 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4193                 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4194                 break;
4195         case MAX_MULTICAST_FILTERS:
4196                 adapter->max_multicast_filters =
4197                     be64_to_cpu(crq->query_capability.number);
4198                 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4199                            adapter->max_multicast_filters);
4200                 break;
4201         case VLAN_HEADER_INSERTION:
4202                 adapter->vlan_header_insertion =
4203                     be64_to_cpu(crq->query_capability.number);
4204                 if (adapter->vlan_header_insertion)
4205                         netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4206                 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4207                            adapter->vlan_header_insertion);
4208                 break;
4209         case RX_VLAN_HEADER_INSERTION:
4210                 adapter->rx_vlan_header_insertion =
4211                     be64_to_cpu(crq->query_capability.number);
4212                 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4213                            adapter->rx_vlan_header_insertion);
4214                 break;
4215         case MAX_TX_SG_ENTRIES:
4216                 adapter->max_tx_sg_entries =
4217                     be64_to_cpu(crq->query_capability.number);
4218                 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4219                            adapter->max_tx_sg_entries);
4220                 break;
4221         case RX_SG_SUPPORTED:
4222                 adapter->rx_sg_supported =
4223                     be64_to_cpu(crq->query_capability.number);
4224                 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4225                            adapter->rx_sg_supported);
4226                 break;
4227         case OPT_TX_COMP_SUB_QUEUES:
4228                 adapter->opt_tx_comp_sub_queues =
4229                     be64_to_cpu(crq->query_capability.number);
4230                 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4231                            adapter->opt_tx_comp_sub_queues);
4232                 break;
4233         case OPT_RX_COMP_QUEUES:
4234                 adapter->opt_rx_comp_queues =
4235                     be64_to_cpu(crq->query_capability.number);
4236                 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4237                            adapter->opt_rx_comp_queues);
4238                 break;
4239         case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4240                 adapter->opt_rx_bufadd_q_per_rx_comp_q =
4241                     be64_to_cpu(crq->query_capability.number);
4242                 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4243                            adapter->opt_rx_bufadd_q_per_rx_comp_q);
4244                 break;
4245         case OPT_TX_ENTRIES_PER_SUBCRQ:
4246                 adapter->opt_tx_entries_per_subcrq =
4247                     be64_to_cpu(crq->query_capability.number);
4248                 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4249                            adapter->opt_tx_entries_per_subcrq);
4250                 break;
4251         case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4252                 adapter->opt_rxba_entries_per_subcrq =
4253                     be64_to_cpu(crq->query_capability.number);
4254                 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4255                            adapter->opt_rxba_entries_per_subcrq);
4256                 break;
4257         case TX_RX_DESC_REQ:
4258                 adapter->tx_rx_desc_req = crq->query_capability.number;
4259                 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4260                            adapter->tx_rx_desc_req);
4261                 break;
4262
4263         default:
4264                 netdev_err(netdev, "Got invalid cap rsp %d\n",
4265                            crq->query_capability.capability);
4266         }
4267
4268 out:
4269         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4270                 adapter->wait_capability = false;
4271                 ibmvnic_send_req_caps(adapter, 0);
4272         }
4273 }
4274
4275 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4276                                struct ibmvnic_adapter *adapter)
4277 {
4278         struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4279         struct net_device *netdev = adapter->netdev;
4280         struct device *dev = &adapter->vdev->dev;
4281         u64 *u64_crq = (u64 *)crq;
4282         long rc;
4283
4284         netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4285                    (unsigned long int)cpu_to_be64(u64_crq[0]),
4286                    (unsigned long int)cpu_to_be64(u64_crq[1]));
4287         switch (gen_crq->first) {
4288         case IBMVNIC_CRQ_INIT_RSP:
4289                 switch (gen_crq->cmd) {
4290                 case IBMVNIC_CRQ_INIT:
4291                         dev_info(dev, "Partner initialized\n");
4292                         adapter->from_passive_init = true;
4293                         adapter->failover_pending = false;
4294                         if (!completion_done(&adapter->init_done)) {
4295                                 complete(&adapter->init_done);
4296                                 adapter->init_done_rc = -EIO;
4297                         }
4298                         ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4299                         break;
4300                 case IBMVNIC_CRQ_INIT_COMPLETE:
4301                         dev_info(dev, "Partner initialization complete\n");
4302                         adapter->crq.active = true;
4303                         send_version_xchg(adapter);
4304                         break;
4305                 default:
4306                         dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4307                 }
4308                 return;
4309         case IBMVNIC_CRQ_XPORT_EVENT:
4310                 netif_carrier_off(netdev);
4311                 adapter->crq.active = false;
4312                 if (adapter->resetting)
4313                         adapter->force_reset_recovery = true;
4314                 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4315                         dev_info(dev, "Migrated, re-enabling adapter\n");
4316                         ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4317                 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4318                         dev_info(dev, "Backing device failover detected\n");
4319                         adapter->failover_pending = true;
4320                 } else {
4321                         /* The adapter lost the connection */
4322                         dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4323                                 gen_crq->cmd);
4324                         ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4325                 }
4326                 return;
4327         case IBMVNIC_CRQ_CMD_RSP:
4328                 break;
4329         default:
4330                 dev_err(dev, "Got an invalid msg type 0x%02x\n",
4331                         gen_crq->first);
4332                 return;
4333         }
4334
4335         switch (gen_crq->cmd) {
4336         case VERSION_EXCHANGE_RSP:
4337                 rc = crq->version_exchange_rsp.rc.code;
4338                 if (rc) {
4339                         dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4340                         break;
4341                 }
4342                 dev_info(dev, "Partner protocol version is %d\n",
4343                          crq->version_exchange_rsp.version);
4344                 if (be16_to_cpu(crq->version_exchange_rsp.version) <
4345                     ibmvnic_version)
4346                         ibmvnic_version =
4347                             be16_to_cpu(crq->version_exchange_rsp.version);
4348                 send_cap_queries(adapter);
4349                 break;
4350         case QUERY_CAPABILITY_RSP:
4351                 handle_query_cap_rsp(crq, adapter);
4352                 break;
4353         case QUERY_MAP_RSP:
4354                 handle_query_map_rsp(crq, adapter);
4355                 break;
4356         case REQUEST_MAP_RSP:
4357                 adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4358                 complete(&adapter->fw_done);
4359                 break;
4360         case REQUEST_UNMAP_RSP:
4361                 handle_request_unmap_rsp(crq, adapter);
4362                 break;
4363         case REQUEST_CAPABILITY_RSP:
4364                 handle_request_cap_rsp(crq, adapter);
4365                 break;
4366         case LOGIN_RSP:
4367                 netdev_dbg(netdev, "Got Login Response\n");
4368                 handle_login_rsp(crq, adapter);
4369                 break;
4370         case LOGICAL_LINK_STATE_RSP:
4371                 netdev_dbg(netdev,
4372                            "Got Logical Link State Response, state: %d rc: %d\n",
4373                            crq->logical_link_state_rsp.link_state,
4374                            crq->logical_link_state_rsp.rc.code);
4375                 adapter->logical_link_state =
4376                     crq->logical_link_state_rsp.link_state;
4377                 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4378                 complete(&adapter->init_done);
4379                 break;
4380         case LINK_STATE_INDICATION:
4381                 netdev_dbg(netdev, "Got Logical Link State Indication\n");
4382                 adapter->phys_link_state =
4383                     crq->link_state_indication.phys_link_state;
4384                 adapter->logical_link_state =
4385                     crq->link_state_indication.logical_link_state;
4386                 break;
4387         case CHANGE_MAC_ADDR_RSP:
4388                 netdev_dbg(netdev, "Got MAC address change Response\n");
4389                 adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4390                 break;
4391         case ERROR_INDICATION:
4392                 netdev_dbg(netdev, "Got Error Indication\n");
4393                 handle_error_indication(crq, adapter);
4394                 break;
4395         case REQUEST_STATISTICS_RSP:
4396                 netdev_dbg(netdev, "Got Statistics Response\n");
4397                 complete(&adapter->stats_done);
4398                 break;
4399         case QUERY_IP_OFFLOAD_RSP:
4400                 netdev_dbg(netdev, "Got Query IP offload Response\n");
4401                 handle_query_ip_offload_rsp(adapter);
4402                 break;
4403         case MULTICAST_CTRL_RSP:
4404                 netdev_dbg(netdev, "Got multicast control Response\n");
4405                 break;
4406         case CONTROL_IP_OFFLOAD_RSP:
4407                 netdev_dbg(netdev, "Got Control IP offload Response\n");
4408                 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4409                                  sizeof(adapter->ip_offload_ctrl),
4410                                  DMA_TO_DEVICE);
4411                 complete(&adapter->init_done);
4412                 break;
4413         case COLLECT_FW_TRACE_RSP:
4414                 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4415                 complete(&adapter->fw_done);
4416                 break;
4417         case GET_VPD_SIZE_RSP:
4418                 handle_vpd_size_rsp(crq, adapter);
4419                 break;
4420         case GET_VPD_RSP:
4421                 handle_vpd_rsp(crq, adapter);
4422                 break;
4423         default:
4424                 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4425                            gen_crq->cmd);
4426         }
4427 }
4428
4429 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4430 {
4431         struct ibmvnic_adapter *adapter = instance;
4432
4433         tasklet_schedule(&adapter->tasklet);
4434         return IRQ_HANDLED;
4435 }
4436
4437 static void ibmvnic_tasklet(void *data)
4438 {
4439         struct ibmvnic_adapter *adapter = data;
4440         struct ibmvnic_crq_queue *queue = &adapter->crq;
4441         union ibmvnic_crq *crq;
4442         unsigned long flags;
4443         bool done = false;
4444
4445         spin_lock_irqsave(&queue->lock, flags);
4446         while (!done) {
4447                 /* Pull all the valid messages off the CRQ */
4448                 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4449                         ibmvnic_handle_crq(crq, adapter);
4450                         crq->generic.first = 0;
4451                 }
4452
4453                 /* remain in tasklet until all
4454                  * capabilities responses are received
4455                  */
4456                 if (!adapter->wait_capability)
4457                         done = true;
4458         }
4459         /* if capabilities CRQ's were sent in this tasklet, the following
4460          * tasklet must wait until all responses are received
4461          */
4462         if (atomic_read(&adapter->running_cap_crqs) != 0)
4463                 adapter->wait_capability = true;
4464         spin_unlock_irqrestore(&queue->lock, flags);
4465 }
4466
4467 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4468 {
4469         struct vio_dev *vdev = adapter->vdev;
4470         int rc;
4471
4472         do {
4473                 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4474         } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4475
4476         if (rc)
4477                 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4478
4479         return rc;
4480 }
4481
4482 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4483 {
4484         struct ibmvnic_crq_queue *crq = &adapter->crq;
4485         struct device *dev = &adapter->vdev->dev;
4486         struct vio_dev *vdev = adapter->vdev;
4487         int rc;
4488
4489         /* Close the CRQ */
4490         do {
4491                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4492         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4493
4494         /* Clean out the queue */
4495         memset(crq->msgs, 0, PAGE_SIZE);
4496         crq->cur = 0;
4497         crq->active = false;
4498
4499         /* And re-open it again */
4500         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4501                                 crq->msg_token, PAGE_SIZE);
4502
4503         if (rc == H_CLOSED)
4504                 /* Adapter is good, but other end is not ready */
4505                 dev_warn(dev, "Partner adapter not ready\n");
4506         else if (rc != 0)
4507                 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4508
4509         return rc;
4510 }
4511
4512 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4513 {
4514         struct ibmvnic_crq_queue *crq = &adapter->crq;
4515         struct vio_dev *vdev = adapter->vdev;
4516         long rc;
4517
4518         if (!crq->msgs)
4519                 return;
4520
4521         netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4522         free_irq(vdev->irq, adapter);
4523         tasklet_kill(&adapter->tasklet);
4524         do {
4525                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4526         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4527
4528         dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4529                          DMA_BIDIRECTIONAL);
4530         free_page((unsigned long)crq->msgs);
4531         crq->msgs = NULL;
4532         crq->active = false;
4533 }
4534
4535 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4536 {
4537         struct ibmvnic_crq_queue *crq = &adapter->crq;
4538         struct device *dev = &adapter->vdev->dev;
4539         struct vio_dev *vdev = adapter->vdev;
4540         int rc, retrc = -ENOMEM;
4541
4542         if (crq->msgs)
4543                 return 0;
4544
4545         crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4546         /* Should we allocate more than one page? */
4547
4548         if (!crq->msgs)
4549                 return -ENOMEM;
4550
4551         crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4552         crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4553                                         DMA_BIDIRECTIONAL);
4554         if (dma_mapping_error(dev, crq->msg_token))
4555                 goto map_failed;
4556
4557         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4558                                 crq->msg_token, PAGE_SIZE);
4559
4560         if (rc == H_RESOURCE)
4561                 /* maybe kexecing and resource is busy. try a reset */
4562                 rc = ibmvnic_reset_crq(adapter);
4563         retrc = rc;
4564
4565         if (rc == H_CLOSED) {
4566                 dev_warn(dev, "Partner adapter not ready\n");
4567         } else if (rc) {
4568                 dev_warn(dev, "Error %d opening adapter\n", rc);
4569                 goto reg_crq_failed;
4570         }
4571
4572         retrc = 0;
4573
4574         tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4575                      (unsigned long)adapter);
4576
4577         netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4578         rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, IBMVNIC_NAME,
4579                          adapter);
4580         if (rc) {
4581                 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4582                         vdev->irq, rc);
4583                 goto req_irq_failed;
4584         }
4585
4586         rc = vio_enable_interrupts(vdev);
4587         if (rc) {
4588                 dev_err(dev, "Error %d enabling interrupts\n", rc);
4589                 goto req_irq_failed;
4590         }
4591
4592         crq->cur = 0;
4593         spin_lock_init(&crq->lock);
4594
4595         return retrc;
4596
4597 req_irq_failed:
4598         tasklet_kill(&adapter->tasklet);
4599         do {
4600                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4601         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4602 reg_crq_failed:
4603         dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4604 map_failed:
4605         free_page((unsigned long)crq->msgs);
4606         crq->msgs = NULL;
4607         return retrc;
4608 }
4609
4610 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter)
4611 {
4612         struct device *dev = &adapter->vdev->dev;
4613         unsigned long timeout = msecs_to_jiffies(30000);
4614         u64 old_num_rx_queues, old_num_tx_queues;
4615         int rc;
4616
4617         adapter->from_passive_init = false;
4618
4619         old_num_rx_queues = adapter->req_rx_queues;
4620         old_num_tx_queues = adapter->req_tx_queues;
4621
4622         init_completion(&adapter->init_done);
4623         adapter->init_done_rc = 0;
4624         ibmvnic_send_crq_init(adapter);
4625         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4626                 dev_err(dev, "Initialization sequence timed out\n");
4627                 return -1;
4628         }
4629
4630         if (adapter->init_done_rc) {
4631                 release_crq_queue(adapter);
4632                 return adapter->init_done_rc;
4633         }
4634
4635         if (adapter->from_passive_init) {
4636                 adapter->state = VNIC_OPEN;
4637                 adapter->from_passive_init = false;
4638                 return -1;
4639         }
4640
4641         if (adapter->resetting && !adapter->wait_for_reset &&
4642             adapter->reset_reason != VNIC_RESET_MOBILITY) {
4643                 if (adapter->req_rx_queues != old_num_rx_queues ||
4644                     adapter->req_tx_queues != old_num_tx_queues) {
4645                         release_sub_crqs(adapter, 0);
4646                         rc = init_sub_crqs(adapter);
4647                 } else {
4648                         rc = reset_sub_crq_queues(adapter);
4649                 }
4650         } else {
4651                 rc = init_sub_crqs(adapter);
4652         }
4653
4654         if (rc) {
4655                 dev_err(dev, "Initialization of sub crqs failed\n");
4656                 release_crq_queue(adapter);
4657                 return rc;
4658         }
4659
4660         rc = init_sub_crq_irqs(adapter);
4661         if (rc) {
4662                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4663                 release_crq_queue(adapter);
4664         }
4665
4666         return rc;
4667 }
4668
4669 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4670 {
4671         struct device *dev = &adapter->vdev->dev;
4672         unsigned long timeout = msecs_to_jiffies(30000);
4673         int rc;
4674
4675         adapter->from_passive_init = false;
4676
4677         init_completion(&adapter->init_done);
4678         adapter->init_done_rc = 0;
4679         ibmvnic_send_crq_init(adapter);
4680         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4681                 dev_err(dev, "Initialization sequence timed out\n");
4682                 return -1;
4683         }
4684
4685         if (adapter->init_done_rc) {
4686                 release_crq_queue(adapter);
4687                 return adapter->init_done_rc;
4688         }
4689
4690         if (adapter->from_passive_init) {
4691                 adapter->state = VNIC_OPEN;
4692                 adapter->from_passive_init = false;
4693                 return -1;
4694         }
4695
4696         rc = init_sub_crqs(adapter);
4697         if (rc) {
4698                 dev_err(dev, "Initialization of sub crqs failed\n");
4699                 release_crq_queue(adapter);
4700                 return rc;
4701         }
4702
4703         rc = init_sub_crq_irqs(adapter);
4704         if (rc) {
4705                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4706                 release_crq_queue(adapter);
4707         }
4708
4709         return rc;
4710 }
4711
4712 static struct device_attribute dev_attr_failover;
4713
4714 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4715 {
4716         struct ibmvnic_adapter *adapter;
4717         struct net_device *netdev;
4718         unsigned char *mac_addr_p;
4719         int rc;
4720
4721         dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4722                 dev->unit_address);
4723
4724         mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4725                                                         VETH_MAC_ADDR, NULL);
4726         if (!mac_addr_p) {
4727                 dev_err(&dev->dev,
4728                         "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4729                         __FILE__, __LINE__);
4730                 return 0;
4731         }
4732
4733         netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4734                                    IBMVNIC_MAX_QUEUES);
4735         if (!netdev)
4736                 return -ENOMEM;
4737
4738         adapter = netdev_priv(netdev);
4739         adapter->state = VNIC_PROBING;
4740         dev_set_drvdata(&dev->dev, netdev);
4741         adapter->vdev = dev;
4742         adapter->netdev = netdev;
4743
4744         ether_addr_copy(adapter->mac_addr, mac_addr_p);
4745         ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4746         netdev->irq = dev->irq;
4747         netdev->netdev_ops = &ibmvnic_netdev_ops;
4748         netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4749         SET_NETDEV_DEV(netdev, &dev->dev);
4750
4751         spin_lock_init(&adapter->stats_lock);
4752
4753         INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4754         INIT_LIST_HEAD(&adapter->rwi_list);
4755         mutex_init(&adapter->rwi_lock);
4756         adapter->resetting = false;
4757
4758         adapter->mac_change_pending = false;
4759
4760         do {
4761                 rc = init_crq_queue(adapter);
4762                 if (rc) {
4763                         dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
4764                                 rc);
4765                         goto ibmvnic_init_fail;
4766                 }
4767
4768                 rc = ibmvnic_init(adapter);
4769                 if (rc && rc != EAGAIN)
4770                         goto ibmvnic_init_fail;
4771         } while (rc == EAGAIN);
4772
4773         rc = init_stats_buffers(adapter);
4774         if (rc)
4775                 goto ibmvnic_init_fail;
4776
4777         rc = init_stats_token(adapter);
4778         if (rc)
4779                 goto ibmvnic_stats_fail;
4780
4781         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4782         netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4783         netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4784
4785         rc = device_create_file(&dev->dev, &dev_attr_failover);
4786         if (rc)
4787                 goto ibmvnic_dev_file_err;
4788
4789         netif_carrier_off(netdev);
4790         rc = register_netdev(netdev);
4791         if (rc) {
4792                 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4793                 goto ibmvnic_register_fail;
4794         }
4795         dev_info(&dev->dev, "ibmvnic registered\n");
4796
4797         adapter->state = VNIC_PROBED;
4798
4799         adapter->wait_for_reset = false;
4800
4801         return 0;
4802
4803 ibmvnic_register_fail:
4804         device_remove_file(&dev->dev, &dev_attr_failover);
4805
4806 ibmvnic_dev_file_err:
4807         release_stats_token(adapter);
4808
4809 ibmvnic_stats_fail:
4810         release_stats_buffers(adapter);
4811
4812 ibmvnic_init_fail:
4813         release_sub_crqs(adapter, 1);
4814         release_crq_queue(adapter);
4815         free_netdev(netdev);
4816
4817         return rc;
4818 }
4819
4820 static int ibmvnic_remove(struct vio_dev *dev)
4821 {
4822         struct net_device *netdev = dev_get_drvdata(&dev->dev);
4823         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4824
4825         adapter->state = VNIC_REMOVING;
4826         rtnl_lock();
4827         unregister_netdevice(netdev);
4828
4829         release_resources(adapter);
4830         release_sub_crqs(adapter, 1);
4831         release_crq_queue(adapter);
4832
4833         release_stats_token(adapter);
4834         release_stats_buffers(adapter);
4835
4836         adapter->state = VNIC_REMOVED;
4837
4838         rtnl_unlock();
4839         device_remove_file(&dev->dev, &dev_attr_failover);
4840         free_netdev(netdev);
4841         dev_set_drvdata(&dev->dev, NULL);
4842
4843         return 0;
4844 }
4845
4846 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4847                               const char *buf, size_t count)
4848 {
4849         struct net_device *netdev = dev_get_drvdata(dev);
4850         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4851         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4852         __be64 session_token;
4853         long rc;
4854
4855         if (!sysfs_streq(buf, "1"))
4856                 return -EINVAL;
4857
4858         rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4859                          H_GET_SESSION_TOKEN, 0, 0, 0);
4860         if (rc) {
4861                 netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4862                            rc);
4863                 return -EINVAL;
4864         }
4865
4866         session_token = (__be64)retbuf[0];
4867         netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4868                    be64_to_cpu(session_token));
4869         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4870                                 H_SESSION_ERR_DETECTED, session_token, 0, 0);
4871         if (rc) {
4872                 netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4873                            rc);
4874                 return -EINVAL;
4875         }
4876
4877         return count;
4878 }
4879
4880 static DEVICE_ATTR_WO(failover);
4881
4882 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4883 {
4884         struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4885         struct ibmvnic_adapter *adapter;
4886         struct iommu_table *tbl;
4887         unsigned long ret = 0;
4888         int i;
4889
4890         tbl = get_iommu_table_base(&vdev->dev);
4891
4892         /* netdev inits at probe time along with the structures we need below*/
4893         if (!netdev)
4894                 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4895
4896         adapter = netdev_priv(netdev);
4897
4898         ret += PAGE_SIZE; /* the crq message queue */
4899         ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4900
4901         for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4902                 ret += 4 * PAGE_SIZE; /* the scrq message queue */
4903
4904         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4905              i++)
4906                 ret += adapter->rx_pool[i].size *
4907                     IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4908
4909         return ret;
4910 }
4911
4912 static int ibmvnic_resume(struct device *dev)
4913 {
4914         struct net_device *netdev = dev_get_drvdata(dev);
4915         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4916
4917         if (adapter->state != VNIC_OPEN)
4918                 return 0;
4919
4920         tasklet_schedule(&adapter->tasklet);
4921
4922         return 0;
4923 }
4924
4925 static const struct vio_device_id ibmvnic_device_table[] = {
4926         {"network", "IBM,vnic"},
4927         {"", "" }
4928 };
4929 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
4930
4931 static const struct dev_pm_ops ibmvnic_pm_ops = {
4932         .resume = ibmvnic_resume
4933 };
4934
4935 static struct vio_driver ibmvnic_driver = {
4936         .id_table       = ibmvnic_device_table,
4937         .probe          = ibmvnic_probe,
4938         .remove         = ibmvnic_remove,
4939         .get_desired_dma = ibmvnic_get_desired_dma,
4940         .name           = ibmvnic_driver_name,
4941         .pm             = &ibmvnic_pm_ops,
4942 };
4943
4944 /* module functions */
4945 static int __init ibmvnic_module_init(void)
4946 {
4947         pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
4948                 IBMVNIC_DRIVER_VERSION);
4949
4950         return vio_register_driver(&ibmvnic_driver);
4951 }
4952
4953 static void __exit ibmvnic_module_exit(void)
4954 {
4955         vio_unregister_driver(&ibmvnic_driver);
4956 }
4957
4958 module_init(ibmvnic_module_init);
4959 module_exit(ibmvnic_module_exit);